System and Method for Stereoscopic Image Acquisition

S. Ristanto1, 2, W. Nugroho1 , E. Sulistya1 and G. B. Suparta1, a)

1
Department of Physics, Faculty of Mathematics and Natural Sciences (FMIPA), Universitas Gadjah Mada,
Yogyakarta 55281, Indonesia
2
Department of Physical Education, Faculty of Mathematics Education, Natural Science and
Information Technology, Universitas PGRI Semarang, Semarang 50232, Indonesia
a)
Corresponding author: gbsuparta@ugm.ac.id

Abstract. This paper discusses stereoscopic images gain using library in OpenCV on Python platform.
The simple stereo camera system that can provide stereoscopic images in both as a picture (PNG) or as a video (AVI) is
developed. The system comprises a stereo camera, a USB connector, a tripod, and a notebook with programming language
of Python 3.7, along with the free library of OpenCV 4.0 installed inside the notebook. The system has successfully
provided both images and videos. However, the spatial resolution and the frame rate depend to the specifications of the
stereo camera being used. The camera resolution of 1280 x 360 with a frame rate of 5 fps is obtained based on the stereo
camera specifications in this study. The disparity of the stereo camera as a function of distance of the camera to the object
of interest is measured. From the experimental results, the p constant is 1059.4, q constant is -0.527 over a distance of
10–100 cm. It shows that farther the object from the camera, the disparity decreases. This system will be used for
performing physical measurement in the future.

Keywords: Stereo camera, image acquisition, disparity, 3-D physical measurement, image analysis

INTRODUCTION

Stereo camera is a two-camera system that has a slightly different angles and can capture two images
simultaneously. In Cartesian coordinate, the two cameras are in the same vertical position, but different horizontal
positions. By utilizing the different perspectives between two cameras, a 3-D image can be reconstructed [1, 2].
The stereo camera is widely chosen because of its low cost, non-contact, flexible, and easy to integrate with other
systems [3].
Stereo camera has been used during manufacturing processes and quality control process in many industries. Stereo
camera also used for various fields such as measuring the density of fish at sea [4, 5], biomass ecosystems at sea [6],
speed Hump detection [7], robotics [8], and measuring deformation [9]. Moreover, to improve image quality, stereo
camera acquisitions can be added to an active device in the form of a binary image using an LCD projector [10].
The way the stereo camera works may become an interesting concern for physical measurement, especially in
3-D physical measurement. Meanwhile, the accuracy of a stereo camera calibration can be improved by the use of
lasers illumination [11]. Continuous calibration is a necessity when doing measurement of moving objects [12].
In some cases, tracking accuracy of stereo cameras is also important [13, 14]. This paper deals with a preliminary
study on how a stereo camera can be used to provide images that eventually their images can be analyzed and used as
physical measurement tool. In this paper, the meaning of disparity is explored.

Proceedings of the 6th International Symposium on Current Progress in Mathematics and Sciences 2020 (ISCPMS 2020)
AIP Conf. Proc. 2374, 020014-1–020014-7; https://doi.org/10.1063/5.0058929
Published by AIP Publishing. 978-0-7354-4113-2/$30.00

020014-1
 METHODOLOGY
This study aims to create a stereo camera system that can produce stereoscopic images using library provided by
OpenCV and it runs under Python platform. The stereoscopic image can be saved as image and video. OpenCV is a
cross-platform open source library that provides various modules for experimentation and computer vision
applications [15]. Meanwhile, Python is a high-level, open source and cross-platform programming language that is
easy to learn [16]. The description of the system and its method is described below.

Hardware Design

The system comprises of a built-in stereo camera, a USB cable, a tripod, and a notebook. The stereo camera setup
is shown in Fig. 1. The two cameras are mounted in parallel in the horizontal direction as shown in Fig. 1a.
The distance of the two cameras is 6 cm, like the average distance of the human eyes. The specifications for both
cameras are the same. The stereo camera is attached to the tripod as shown in Fig. 1b, in order to adjust the height and
angle of the camera. Figure 1c shows complete system configuration.

Software Design

The software used is a free OpenCV 4.0, which is an optimized computer vision library specifically designed for
real-time application. It is run under the Python 3.7 platform programming language. Programming algorithms are
used as follows. The software calls all required modules, then it read the address of the stereo camera. After that,
it determines the camera resolution, and specifies the video recording frames per second (fps), the type of image
produced in color, and the image format (AVI). Next, it shows stereo image, then records the images. While it captures
the images, it also calculates the disparity of the camera. At last, it saves all images, video and the disparity as data
set.
The module numpy was used for analyzing video frames, while the module cv2 was used for image processing
and manipulation. As an example, the command of “cv2.video capture (1)” provided in module will access first
external camera connected to the notebook. The camera resolution was selected following the specifications of the
stereo camera. The image is saved as color image and also as video at 5 fps.

(a)

(b) (c)
FIGURE 1. Stereo camera set-up (a) Front view, (b) Side view, and (c) whole system configuration.

020014-2
 Stereo Camera Performance Test
The stereo camera has produced two images of an object with different perspectives in the horizontal direction.
This difference perspective is defined as a disparity. Furthermore, the perception of 3-D images is generally perceived
based on this disparity, just like the way of human eye works. Evaluation on the stereo camera performance relies on
its disparity profile. The profile disparity as a function of the distance between the object and the stereo camera must
be collected.
From pinhole model [17], the disparity (x) can be calculated using equation below

(1)

where b is a baseline, f is a focal length of cameras, and D is a distance object to cameras.

To accommodate experiment result, Equation 1 can be expressed in power function form below

(2)

where p =bf are constants that depending on the specifications of stereo cameras used. The q's value is obtained through
the experiment. Theoretically, q's value is -1. The D in this study varies 10–100 cm.
To measure disparity from stereoscopic images it can be use equation below,

(3)

where x1 and x2 are horizontal object position on right and left images [17].
The specifications of stereo camera using in this study are showed at Table 1.

RESULTS AND DISCUSSION

Resolution Testing

In the first step, the resolution of the stereo camera used in this system was tested. Therefore, the stereo camera
register was called up by using the module program “cv2.video capture (1)”. Without a definition of the image
resolution entered, the system will fail to acquire the stereoscopic image. The camera would be able to provide a stereo
image when the aspect ratio of the resolution was 32:9. Therefore, the image resolutions of 2560x720, 1280x360,
1920x540, and 640x180 were working well. So, there were four options of image resolution could be used. The stereo
image is essentially two images that are put side by side. For a stereo image of 1280x360, it comprises of two images
at a resolution of 640x360.

TABLE 1. Camera stereo specifications

Specifications Value
Baseline (b) 6 cm
Maximum Resolution 2560 x 720 pixel
Focal length ( f ) 600 pixel

020014-3
 Frame Rate Testing
Next test is determining the best frame rate for the image acquisition. At the first trial, a frame rate of 30 fps was
used. The results showed that the video was good, but it was too fast. Then the frame rate was gradually decreased by
5 from 25 fps, 20 fps, 15 fps, 10 fps and 5 fps. The results showed that the frame rate of 5 fps was the best corresponds
to the real tempo.

Stereoscopic Images Testing

Based on the results of previous tests, the stereoscopic image resolution of 1280x360 and a frame rate of 5 fps
were selected. An example of the stereoscopic image produced in this study is shown in Fig. 2. The image shows that
the left image and the right image is shifted in horizontal direction. This horizontal shift is further defined as a disparity.
Based on this disparity, the algorithm can obtain stereoscopic images. The algorithm is implemented as a software
written in Python using OpenCV modules.
When the object distance to the camera changed, the disparity also changed. By acquiring the disparity as a function
of the object distance to the camera, the disparity profile obtained. The disparity profile can be used to obtain 3-D
images by applying the triangulation concept [18].
Examples of stereoscopic images acquired at a distance of 20cm to 50 cm is shown in Fig. 3. The object was a
board that had a chess motif. The stereo camera stood perpendicular relative to the floor at a height of 50 cm from the
floor. The camera was kept at fixed position, while the distance of the object from the stereo camera changed.

Disparity

The disparity profile as a function of the object distance to the camera in the range of 10–100 cm is shown in
Fig. 4. It presents graphs of the disparity from experimental results and theoretical approach. Both graphs agree that
the farther the object from the camera, the shorter the disparity. Then the curve-fitting based on power function is
applied. The experimental results provide disparity profile of y = 1096.4x-0.527 with a regression value R2 of 0.9787.
From these result, the p constant is 1059.4, q constant is -0.527 based on Equation 2. This power function of
experimental disparity is more sloping than the theoretical disparity profile of y = 3590.4x-1 with a regression value
R2 of 1.0. This difference is caused by an alignment error [17]. However, this difference does not affect the quality of
subsequent data processing. The theoretical approach is only to help construct the disparity equation that is to seek
through experiments. The data used for the next process is the disparity equation obtained through experiments.

Anaglyph Image

Anaglyph image is another presentation for a stereoscopic image. The module for converting the stereoscopic
image as an anaglyph image was developed. Using the module, the stereoscopic image was divided into two parts,
left image (L) and right image (R). For the stereoscopic image of 1280x360 in resolution, there were two images, each
had a resolution of 640x360. Then, the left image is converted to be a blue image, while the right image is converted
to be a red image. After that, both images were superimposed to yield single image. Figure 5 shows the anaglyph
images for the stereoscopic images in Fig. 3. The anaglyph image can be examined using 3-D glasses.

FIGURE 2. Stereoscopic image display using a stereo camera at image resolution of 1280 x 360.

020014-4
 (a)

(b)

(c)

(d)
FIGURE 3. The stereoscopic image acquired at a distance of (a) 20 cm, (b) 30 cm, (c) 40 cm and (d) 50 cm from the camera.
The stereoscopic images (a) – (d) represent a combination of associated left image and right image, where there is a shift
in the horizontal direction between both images. Each shift depends to the distance of the object from the camera.

020014-5
 FIGURE 4. The disparity profile as a function of the object distance to the camera, resulted from theory and experiment.

(a) (b)

(c) (d)
FIGURE 5. The anaglyph images of the stereoscopic image at a distance of (a) 20 cm, (b) 30 cm,
(c) 40 cm and (d) 50 cm from the camera in Fig. 3.

CONCLUSION

This experience showed that a stereo camera system and its acquisition method using OpenCV could be built
successfully. It can obtain a stereoscopic image with a resolution of 1280x360 and a frame rate of 5 fps, at least defined
by the stereo camera specifications. It can provide images as video and picture. The stereoscopic image can be
transformed as an anaglyph image. The disparity profile as a function of the camera distance to the object yielded
from experiment shows similar profile with the theory, although it has different slope. From the experimental results,

020014-6
the p constant is 1059.4, q constant is -0.527 over a distance of 10–100 cm. We conclude that as the object getting
further from the camera, the disparity decreases.

ACKNOWLEDGMENTS

We would like to express our gratitude to Deputi Bidang Penguatan Riset dan Pengembangan Kemenristek/BRIN
for providing the research grant of the 2019 Doctoral Dissertation Program and to PT Madeena Karya Indonesia for
facilitating the devices for this research.

REFERENCES

1. H. Kim and J. Lee, Displays 40, 24-34 (2015).

2. S. Hong, A. Ansari, G. Saavedra and M. Martinez-Corral, Opt. Lasers Eng. 103, 46-54 (2018).
3. Y. Chi, L. Yu and B. Pan, Opt. Lasers Eng. 104, pp. 141-148 (2018).
4. K. Williams, N. Lauffenburger, M. -C. Chuang, J. -N. Hwang and R. H. Towler, Methods Oceanogr. 17, 138-152
(2016).
5. K. Williams, C. N. Rooper, A. de Robertis, M. Levine and R. H. Towler, J. Exp. Mar. Biol. Ecol. 508, 21-26
(2018).
6. J. L.Boldt, K. Williams, C. N. Rooper, R. H. Towler and S. Gauthier, Fis. Res. 198, 66-77 (2018).
7. V. S. K. P. Varma, S. Adarsh, K. I. Ramachandran and B. B. Nair, Procedia Comput. Sci. 143, 988-997 (2018).
8. Y. Liu and L. Liu, Rob. Auton. Syst. 101, 34-44 (2018).
9. L. Yu and B. Pan, Opt. Lasers Eng. 95, 17-25 (2017).
10. S. Barone, P. Neri, A. Paoli and A. V. Razionale, Opt. Lasers Eng. 127, 105989 (2020).
11. V. A. Sablina and M. B. Nikiforov, 2016 International Conference on Robotics and Machine Vision. SPIE 10253,
1025302 (2017).
12. G. R. Mueller and H. Wuensche, IEEE 20th International Conference on Intelligent Transportation Systems
(ITSC), Yokohama, 1-6 (2017), doi: 10.1109/ITSC.2017.8317675.
13. J. Ahn, H. Choi, J. Hong and J. Hong, J. Oral Maxillofac. Surg. 77, P1070.e1-1070.e11 (2019).
14. F. Ke, H. Liu, D. Zhao, G. Sun, W. Xu and W. Fang, Optik 204, 164-186 (2020).
15. J. Howse, OpenCV Computer Vision with Python (Packt Publishing Ltd., Birmingham, 2013).
16. B. A. Meier, Python GUI Programming Cookbook 2nd edition (Packt Publishing Ltd., Birmingham, 2017).
17. M. A. Mahammed, A. I. Melhum, and F. A. Kochery, International Journal of Science and Applied Information
Technology (IJSAIT) 2, 05-08 (2013)
18. S. Gai, F. Da and X. Dai, Opt. Lasers Eng. 104, 126-134 (2017).

020014-7
 Vol.11 (2021) No. 6
ISSN: 2088-5334

3D Velocity Measurement of Translational Motion using a Stereo

Camera
Sigit Ristanto a,b, Waskito Nugroho a, Eko Sulistya a, Gede Bayu Suparta a,*
a
Department of Physics, Universitas Gadjah Mada, Yogyakarta, 55281, Indonesia
b
Department of Physics Education, Universitas PGRI Semarang, Semarang, 50232, Indonesia
Corresponding author: *gbsuparta@ugm.ac.id

Abstract— This research aims to create a 3D velocity measurement system using a stereo camera. The 3D velocity in this paper is the
velocity which consists of three velocity components in 3D Cartesian coordinates. Particular attention is focused on translational motion.
The system set consists of a stereo camera and a mini-PC with Python 3.7, and OpenCV 4.0 installed. The measurement method begins
with the selection of the measured object, object detection using template matching, disparity calculation using the triangulation
principle, velocity calculation based on object displacement information and time between frames, and the storage of measurement
results. The measurement system's performance was tested by experimenting with measuring conveyor velocity from forward-looking
and angle-looking directions. The experimental results show that the 3D trajectory of the object can be displayed, the velocity of each
component and the speed as the magnitude of the velocity can be obtained, and so the 3D velocity measurement can be performed. The
camera can be positioned forward-looking or at a certain angle without affecting the measurement results. The measurement of the
speed of the conveyor is 11.6 cm/s with an accuracy of 0.4 cm/s. The results of this study can be applied in the performance inspection
process of conveyors and other industrial equipment that requires speed measurement. In addition, it can also be developed for accident
analysis in transportation systems and practical tools for physics learning.

Keywords— Stereo vision; speed measurement; object tracking; visual odometry; autonomous driving.

Manuscript received 18 Jun. 2021; revised 31 Jul. 2021; accepted 12 Aug. 2021. Date of publication 31 Dec. 2021.
IJASEIT is licensed under a Creative Commons Attribution-Share Alike 4.0 International License.

cameras require that the object's size must be known to obtain

I. INTRODUCTION
Velocity measurements are needed in various fields such
as transportation, robotics, satellite navigation, and
manufacturing. In transportation, velocity measurement is
important for accident analysis [1], [2], odometry [3], [4], and
autonomous driving [5]. In robotics, velocity is one of the
parameters needed in robot navigation [6], [7]. In satellite
navigation, velocity measurement is needed to position the
satellite as accurately as possible in its orbit [8], [9]. In
manufacturing, velocity measurement is carried out as part of
conveyor performance control [10], [11].
The existing velocity measurement methods include Lidar,
Radar, and video/image. Video/image measurement methods
are being developed because they have the potential to
measure speed and identify vehicle numbers [12], [13].
Moreover, video/image speed measurements can be used for
autonomous car control systems [14].
Based on the camera used, there are two methods, single
cameras [15]–[17] and stereo cameras [18]–[20]. Single
the measurement results when a stereo camera uses the
principle of triangulation. In terms of performance, the
measurement accuracy of a stereo camera is potentially
superior to that of a single camera. Stereo cameras can get
more data and produce fewer occlusions than a single camera.
Any occlusion caused by surface geometry blocks projected
light or the camera’s view with a single camera would result in
no data. On the other hand, stereo cameras effectively use three
views of the part to capture more data. Simultaneously,
triangulation of the left camera and triangulation of the right
camera take more data than a single camera, and then stereo
may be obtained with two cameras working together.
Based on the placement of the detector, there are two
methods, namely forward-looking and side-looking [21].
Forward-looking is a measurement where the detector is
placed parallel to the direction of motion, while side-looking
is a measurement where the detector is placed perpendicular
to the direction of motion. Both of these placements have the
advantage of reducing the three-speed components into one-

2482
speed components, making it simpler. However, the
measurement results will be biased if the object's motion does
not match the placement of the camera.
The solution to this problem is to use 3D velocity
measurement, where all three velocity components are
measured simultaneously. Thus, the placement of the detector
is more flexible without changing the programming algorithm.
The detector can not only be placed forward-looking and side-
looking but can be placed angle-looking where the object and
camera form a certain angle.
Research on 3D velocity measurement from angle-looking
is still limited because it involves more variables and is more
complex. However, the advantage of this approach is that the
camera can be placed at various angles so that it is more
flexible in use than forward-looking and side looking.
This paper proposed 3D velocity measurement using a
stereo camera. The detection of objects uses a template
matching algorithm. The measurement process is carried out
by forward-looking as is usually done and then by angle
looking as another alternative. Angle-looking is needed to
show that the 3D velocity has been successfully measured so
that the difference in the placement of the detector affects the
value of the velocity vector component only, but the
magnitude of the velocity is constant. This hypothesis will be
tested with an experiment. The motion measured in this study
is limited to translational motion. Translational motion is the
linear motion of an object accompanied by the displacement
of that object. Examples of translational motion include the Fig. 1 System performance flowchart
trajectory of a car on the highway, the rate of a conveyor belt,
the trajectory of an airplane, a train, and so on. References to The measurement process begins by taking a stereo image
the use of stereo cameras for 3D speed measurement in with 1280 x 480 pixels resolution. The stereo image combines
translational motion are still limited. the left image and the right image, each with a resolution of
640 x 480 pixels.
II. MATERIALS AND METHOD From the resulting stereo image, the object can be
measured manually by hovering the cursor over the object and
A. System Description then pressing the left mouse button. The selected object is
The system used consists of a calibrated stereo camera and automatically used as a template measuring 24x24 pixels. The
a mini-PC with the specifications shown in Table 1. The template represents the object in the real space being
stereo camera is mounted on a tripod connected to the Mini measured. Templates can be changed before taking
PC using a USB cable. measurements or at the time of taking measurements. Users
TABLE I can freely choose objects in the stereo image to be used as
CAMERA AND MINI PC SPECIFICATIONS templates.
Stereo Camera Specifications Next, template matching is applied to the stereo image. The
Focus 3,0 mm (fixed) template matching process is carried out using the library
Sensor ¼ inch contained in OpenCV 4.0. In this paper, the template
Pixel size 3 µm x 3 µm matching process is carried out by dividing the stereo image
Object distance 30 cm – infinity into two equal parts, namely the left image and the right image,
FoV 100º
Image Resolution 2560x720,1280x480,
and then applying template matching to both separately. At
FPS 30 FPS this stage, there are two conditions. If the template on the left
Baseline 6 cm and right images is detected, then proceed to the next process,
Mini PC Specifications but if not, the information "No object detected" is displayed,
Processor Intel(R)Core(TM) i5-8250U CPU@1,6 GHz and the template selection process can be repeated.
1,8 GHz After the object is detected, the next step is to calculate the
RAM 16 GB disparity and 3D position based on the parameters that have
System type 64-bit OS, x64-based processor
Operating system Windows 10
been obtained during calibration. The disparity is the
Software Python 3.7 and OpenCV 4.0 difference in the position of objects on the left and right
images. The disparity is calculated using the pinhole model
The flowchart of the 3D velocity measurement system is approach, as shown in equation (1)
shown in Fig. 1. = − (1)

2483
The 3D position of each component is calculated using
equation (2 – 4)

= (2)

= (3)

= (4)

Next, the 3D velocity component is calculated using

equation (5 – 7),
_ _
= (5) Fig. 2b Angle-looking velocity measurement
!
" _ " _
" = !
(6) The left camera is selected as the reference point (origin).
The X-axis is in the horizontal direction, the Y-axis is in the
# _ # _
= !
(7) vertical direction, and the Z-axis is perpendicular to the stereo
image plane.
The parameter measured includes the components of
velocity, velocity, speed, and speed accuracy. Accuracy was
Then, the 3D velocity vector can be written in the form calculated using the equation mean absolute deviation (MAE)
$= "% & #' & ( (8) [11], as shown in equation (10)
2
While speed as the magnitude of velocity can be calculated /01 = ∑3
782|67 − 6 | (10)
3
using equation (9),
with m as the number of measurements, yi as the ith
) = * )+ , & )- , & ). , (9) measurement, and y as the average of the measurement results.

III. RESULTS AND DISCUSSION

Finally, the disparity data, the object position in the image,
3D position, and 3D velocity are displayed on the stereo A. Measurement Process
image frame. If needed, the data can be saved in the form of
The stereo image display is shown in Fig. 3. Detected
an ‘AVI’ format video and ‘xlsx’ format spreadsheet.
objects are marked with red boxes on the left and right images.
B. Experimental Design The image coordinate axis (x, y) is marked with a white line
with the center of the coordinates (x=0,y=0) located at the
The experiment was carried out on an object placed on a
center of the camera. The x-axis of the image is marked with
conveyor that was constantly moving away from the stereo
a horizontal white line, while the y-axis of the image is
camera, as shown in Fig.2(a-b). To determine the effect of
marked with a vertical white line.
angle, the position of the stereo camera is varied starting from
being in line with the direction of the object's motion as in Fig.
2a (forward-looking) and forming an angle (angle-looking) as
in Fig. 2b. The angle variation used is 30° and 45°.

Fig. 3 An example of a video frame display in the process of measuring speed

The disparity and object position values in the stereo image

Fig. 2a Forward-looking velocity measurement
are shown at the top left, while the 3D position and 3D
velocity values are shown at the bottom left. The 3D Position
and 3D velocity values are displayed as vector components on
each axis.
B. Object Trajectory
The object's trajectory can be obtained through 3D Graphs
of changes in the object's position in the measurement process.
The motion of the translation object will show a straight-line
pattern. The results of measuring changes in object position
in a forward-looking manner are shown in Fig. 4a. Based on
the figure, it appears that the object's trajectory forms a

2484
straight line, changing significantly on the Z-axis with a range
of 35-50 cm, while on the X-axis and Y-axis, it remains the
same.
The results of the angle-looking measurement with an
angle of 30° and 45° to the direction of the object's motion are
shown in Fig. 4a and Fig. 4b. Based on the figure, it appears
that the trajectory of the object is in the form of a straight line,
as in Fig. 4a. The position component on the X-axis and Z-
axis change significantly while the Y position component
remains. However, the Y position component at an angle of
45° is greater than that of an angle of 30°.
Based on Fig. 4(a-c), it can be shown that the object's
trajectory is in the form of a straight line according to the
object's motion on the conveyor. Thus this speed
measurement system has succeeded in showing the 3D
trajectory of the object.
The trajectory of an object is important for visually
demonstrating 3D motion.
Fig. 4c The trajectory of the object when the camera makes an angle of 45°
to the direction of motion

C. Velocity Component, Velocity, and Speed Measurement

An example of a position graph as a function of time for
three variations of camera position is shown in Fig. 5(a-c).
Based on Fig. 5(a-c), it can be seen that the relationship
between position and time on all components of velocity is
linear so that the 3D velocity measurement results are
constant in all variations.
The value of the velocity component on each coordinate
axis is equivalent to the gradient in the equation obtained by
curve fitting a linear equation. In Fig. 5a it can be concluded
that the velocity component on each axis is VX = -0.35 cm/s,
VY = 0.18 cm/s and VZ = 11.8 cm/s.

Based on equation (8), the velocity can be expressed as

$ = 0.35% & 0.18' & 11.8( cm/s.

Fig. 4a The trajectory of the object when the camera is in line with the
direction of motion
The speed can be calculated using equation (9) as

V = * −0.35 ? & −0.18 ? & −11,85 ? =11.8 cm/s.

These results indicate that the forward-looking velocity

component value on the Z-axis is optimum, V = VZ, while VX
and VY are very small.

Fig. 4b The trajectory of the object when the camera makes an angle of 30°
to the direction of motion
Fig. 5a Position component measurement results on Forward-looking

2485
 In the same way, based on Fig. 5b, the velocity component
obtained at 30° looking angle are VX = 5.6 cm/s, VY = 0.15
cm/s, VZ = 10.5 cm/s. The velocity $ = 5.6% & 0.15' &
10.5( cm/s and the speed V = 11.9 cm/s. The angle value can
be found using the equation = tan-1(VX/ VZ)=28°. This value is
2° degrees different from what it should be.

Fig. 6 Speed and its accuracy in various directions

The results of this experiment produce a maximum

Fig. 5b The result of measuring the position component at an angle-looking
30°
In the same way, based on Fig. 5c, the velocity component
obtained at the angle looking 45° is VX = 7.9 cm/s, VY = 0.2
cm/s, VZ = 8.5 cm/s. The velocity $ = 7.9% & 0.2' & 8.5(
cm/s. and the speed V = 11.6 cm/s. The angle value is θ =
Atan-1(VX / VZ) = 43°.
The angle values on the angle-looking 30° and 45° are both
2° apart from what they should be. Thus, the velocity
component can be measured with an angle error of ± 2°.
uncertainty of 0.4 cm/s, which has higher precision than the
conveyor speed measurement using a single camera that
produces an uncertainty of 1 cm/s [11]. Another advantage of
this system is that not only the speed obtained but also 3D
velocity data is obtained so that the measurement process is
more flexible, which can be carried out from a forward-
looking or angle-looking perspective. In the [11]
measurement method, the measurement process is limited to
side-looking. Yang et al. [13] also utilize a stereo camera for
speed measurement. The system is designed to measure all
vehicles' speed in ROI automatically. The difference with the
system developed in this study is that the 3D velocity
measurement is carried out on the desired object only. The
maximum accuracy of speed measurement by Yang et al. [13]
is 3.8%, while it was 3.4% in this study. Thus, the proposed
method was more accurate.

IV. CONCLUSION
The system developed in this study has successfully
measured 3D velocity. The camera position can be placed
forward-looking or angle-looking. Based on the specifications
of the stereo camera used, this system can measure the speed
of 11.6 cm/s with an absolute error of 0.4 cm/s.

NOMENCLATURE
d disparity pixel
xr horizontal position at right image pixel
xl horizontal position at left image pixel
Fig. 5c The result of measuring the position component at an angle-looking
f focus length pixel
45° b baseline cm
X position component on X-axis cm
A. Speed Measurement and its Accuracy Y position component on Y-axis cm
Speed accuracy is calculated using equation 10. The Z position component on Z-axis cm
measurement is carried out five times. The measurement of V velocity cms-1
the speed for the three variations is shown in Fig. 6. The V speed cms-1
results for each variation are forward-looking (11.6 ± 0.4) θ angle between camera to object direction radian
cm/s, angle-looking 30° (11.8 ± 0, 2) cm/s, and the angle- VX velocity component on X-axis cms-1
looking 45° is (11.7 ± 0.3 ) cm/s. Based on Fig. 6, it appears VY velocity component on X-axis cms-1
that the measurement results along with the angle-looking VZ velocity component on X-axis cms-1
errors of 30° and 45° are within the range of forward-looking FPS Frame per Second s-1
error. Thus, it can be stated that the measurement results of i vector unit on X-axis
the three variations are consistent with each other. j vector unit on Y-axis
k vector unit on Z-axis

2486
 ACKNOWLEDGMENT
We are grateful to the Ministry of Research, Technology,
and Higher Education of Indonesia through BPPDN grants
and the 2020-2021 Doctoral Research Grant for funding this
research. Also, we are obliged to PT Madeena Karya
Indonesia for preparing the stereo camera used in this
prototype.
REFERENCES
[1] J. H. Kim, W. T. Oh, J. H. Choi, and J. C. Park, “Reliability
verification of vehicle speed estimate method in forensic videos,”
Forensic Sci. Int., vol. 287, pp. 195–206, 2018.
[2] L. R. Costa, M. S. Rauen, and A. B. Fronza, “Car speed estimation
based on image scale factor,” Forensic Sci. Int., vol. 310, p. 110229,
2020.
[3] S. J. Yoon and T. Kim, “Development of stereo visual odometry based
on photogrammetric feature optimization,” Remote Sens., vol. 11, no.
1, 2019.
[4] E. Jung, N. Yang, and D. Cremers, “Multi-frame GAN: Image
enhancement for stereo visual odometry in low light,” arXiv, no. CoRL
2019, pp. 1–10, 2019.
[5] R. Fan, L. Wang, M. J. Bocus, and I. Pitas, “Computer stereo vision
for autonomous driving,” arXiv, 2020.
[6] L. Li, Y. H. Liu, T. Jiang, K. Wang, and M. Fang, “Adaptive
Trajectory Tracking of Nonholonomic Mobile Robots Using Vision-
Based Position and Velocity Estimation,” IEEE Trans. Cybern., vol.
48, no. 2, pp. 571–582, 2018.
[7] Y. Liu, Y. Gu, J. Li, and X. Zhang, “Robust stereo visual odometry
using improved RANSAC-based methods for mobile robot
localization,” Sensors (Switzerland), vol. 17, no. 10, 2017.
[8] J. Guo et al., “Real-time measurement and estimation of the 3D
geometry and motion parameters for spatially unknown moving
targets,” Aerosp. Sci. Technol., vol. 97, p. 105619, 2020.
[9] D. Ge, D. Wang, Y. Zou, and J. Shi, “Motion and inertial parameter
estimation of non-cooperative target on orbit using stereo vision,” Adv.
Sp. Res., vol. 66, no. 6, pp. 1475–1484, 2020.
[10] Z. Li, F. Zeng, C. Yan, J. Wang, and L. Tang, “Design of belt conveyor
speed control system of ‘internet+,’” IOP Conf. Ser. Mater. Sci. Eng.,
vol. 563, no. 4, 2019.
2487
[11] Y. Gao, T. Qiao, H. Zhang, Y. Yang, Y. Pang, and H. Wei, “A
contactless measuring speed system of belt conveyor based on
machine vision and machine learning,” measurement, vol. 139, pp.
127–133, 2019.
[12] P. Najman and P. Zemcik, “Vehicle Speed Measurement Using Stereo
Camera Pair,” IEEE Trans. Intell. Transp. Syst., pp. 1–9, 2020.
[13] L. Yang, M. Li, X. Song, Z. Xiong, C. Hou, and B. Qu, “Vehicle Speed
Measurement Based on Binocular Stereovision System,” IEEE Access,
vol. 7, pp. 106628–106641, 2019.
[14] A. Safaei, “Adaptive relative velocity estimation algorithm for
autonomous mobile robots using the measurements on acceleration
and relative distance,” Int. J. Adapt. Control Signal Process., vol. 34,
no. 3, pp. 372–388, 2020.
[15] H. Deng, U. Arif, K. Yang, Z. Xi, Q. Quan, and K. Y. Cai, “Global
optical flow-based estimation of velocity for multicopters using
monocular vision in GPS-denied environments,” Optik (Stuttg)., vol.
219, no. March 2019, p. 164923, 2020.
[16] G. Gallego and D. Scaramuzza, “Accurate angular velocity estimation
with an event camera,” IEEE Robot. Autom. Lett., vol. 2, no. 2, pp.
632–639, 2017.
[17] F. Guo and R. Chellappa, “Video Metrology Using a Single Camera,”
IEEE Trans. Pattern Anal. Mach. Intell., vol. 32, no. 7, pp. 1329–1335,
2010.
[18] N. Murmu, B. Chakraborty, and D. Nandi, “Relative velocity
measurement using low cost single camera-based stereo vision system,”
Measurement, vol. 141, pp. 1–11, 2019.
[19] Y. C. Lim, M. Lee, C. H. Lee, S. Kwon, and J. H. Lee, “Improvement
of stereo vision-based position and velocity estimation and tracking
using a stripe-based disparity estimation and inverse perspective map-
based extended Kalman filter,” Opt. Lasers Eng., vol. 48, no. 9, pp.
859–868, 2010.
[20] “A Novel Approach to Measurement of the Transverse Velocity of the
Large-Scale Objects.” [Online]. Available:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7988419/. [Accessed:
30-Apr-2021].
[21] S. L. Jeng, W. H. Chieng, and H. P. Lu, “Estimating speed using a
side-looking single-radar vehicle detector,” IEEE Trans. Intell. Transp.
Syst., vol. 15, no. 2, pp. 607–614, 2014.
Physics Education

PAPER You may also like

- Real-time vision-based grasping randomly
Development of laboratory devices for real-time placed object by low-cost robotic arm
using surf algorithm
measurement of object 3D position using stereo A Beyhan and N G Adar

- (Invited) Bio-Sniffers & Sniff-Cam:

cameras Biofluorometirc Gas Sensor and Imaging
System for Human Volatile Chemicals
Kohji Mitsubayashi
To cite this article: Sigit Ristanto et al 2022 Phys. Educ. 57 025006
- (Invited) Bio-Sniffers & Sniff-Cam:
Biofluorometirc Gas Sensor and Imaging
System for Human Volatile Chemicals
Kohji Mitsubayashi

View the article online for updates and enhancements.

This content was downloaded by joe.murphy@ioppublishing.org from IP address 51.104.190.136 on 27/01/2022 at 15:34
 PAPER
Phys. Educ. 57 (2022) 025006 (12pp) iopscience.org/ped

Development of
laboratory devices for
real-time measurement of
object 3D position using
stereo cameras
Sigit Ristanto1,2, Waskito Nugroho1, Eko Sulistya1
and Gede B Suparta1,∗
1
Department of Physics, Universitas Gadjah Mada, DI Yogyakarta 55281, Indonesia
2
Departement of Physics Education, Universitas PGRI Semarang, Semarang 50232,
Indonesia

E-mail: gbsuparta@ugm.ac.id

Abstract
Measuring the 3D position at any time of a given object in real-time
automatically and well documented to understand a physical phenomenon is
essential. Exploring a stereo camera in developing 3D images is very
intriguing since a 3D image perception generated by a stereo image may be
reprojected back to generate a 3D object position at a specific time. This
research aimed to develop a device and measure the 3D object position in
real-time using a stereo camera. The device was constructed from a stereo
camera, tripod, and a mini-PC. Calibration was carried out for position
measurement in X, Y, and Z directions based on the disparity in the two
images. Then, a simple 3D position measurement was carried out based on
the calibration results. Also, whether the measurement was in real-time was
justified. By applying template matching and triangulation algorithms on
those two images, the object position in the 3D coordinate was calculated
and recorded automatically. The disparity resolution characteristic of the
stereo camera was obtained varied from 132 pixels to 58 pixels for an object
distance to the camera from 30 cm to 70 cm. This setup could measure the
3D object position in real-time with an average delay time of less than 50 ms,
using a notebook and a mini-PC. The 3D position measurement can be
performed in real-time along with automatic documentation. Upon the stereo
camera specifications used in this experiment, the maximum accuracy of the
measurement in X, Y, and Z directions are ∆X = 0.6 cm, ∆Y = 0.2 cm, and
∆Z = 0.8 cm at the measurement range of 30 cm–60 cm. This research is

∗
Author to whom any correspondence should be addressed.

1361-6552/22/025006+12$33.00 1 © 2021 IOP Publishing Ltd

S Ristanto et al
expected to provide new insights in the development of laboratory tools for learning
physics, especially mechanics in schools and colleges.
Keywords: 3D positions, stereo vision, real-time measurement, disparity
Supplementary material for this article is available online
1. Introduction
1.1. Background
Measurement of the 3D object position using a
digital camera helps determine a physical system.
Many inspections, monitoring, robotics, sports
competitions, and accident investigations perform
as 3D images. The 3D measurements using a ste-
reo camera may be performed without direct con-
tact with the object during measurements. The
binocular stereo camera may be developed for
the 1D distance measurement [1, 2] and distance
measurement between objects [3]. Measurements
using binocular stereo cameras have been used
to observe the propagation of distances on hot
objects [4] and the measurement of tube bending
radius [5]. Stereo cameras are also used for meas-
uring the 3D vibrations of gas turbine engines in
aircraft [6]. The accuracy of 3D distance meas-
urement is improved by using an optoelectronic
device based on angular triangulation [7]. Stereo
cameras may be used to track moving objects [8]
and measure the relative velocity of objects [9].
However, the sensitivity, resolution, and speed of
camera technology acquisition continue to evolve
so that the accuracy and reliability of recording
physical symptoms also progressively improve.
With the support of digital image processing,
detecting the position of moving objects from time
to time improves, and then it may take place auto-
matically and in real-time.
There are three popular techniques developed
to detect the position of an object automatically
using a stereo camera, namely background sub-
traction [10–12], optical flow [13–17], and tem-
plate matching [18–21].
The background subtraction method is an
automatic detection method by reducing the cam-
era image relative to the background image. This
method reduces the dynamic background image
[10] or compares the three adjacent images [11].
March 2022
The accuracy improves by applying neural net-
work convolution [12]. This background sub-
traction method uses a simple programming
algorithm. However, this method requires a back-
ground image as a reference. Hence, this method
will be effective by adjusting the background
regardless of the presence of the object.
The optical flow method is a pixel-based
motion detection method [13]. This method is
applied for tracking icebergs using a camera with
a speed of 0.5–4 frames per minute [14]. This
method performs well to detect moving objects
but assumes constant object velocity [15]. Accur-
acy in measuring the object’s position is also
excellent and can run in real-time using Bayesian
Forecast [16]. This method can also work well
for left ventricular tracing to attempt disease dia-
gnosis [17]. However, this method requires ana-
lysis between frames so that it cannot work on a
single frame.
The template matching method compares
two templates using intensity values and gradient
pixels [18]. The accuracy of the template match-
ing method improves by combining it with audio
mapping [19]. Moreover, its constancy improves
by semantic super pixel region [18] and correl-
ation filter-based trackers [20]. This method has
also successfully improved object tracking per-
formance in robotic systems [21].
This paper describes our efforts to develop
a device and method for measuring the 3D pos-
ition of an object in real-time using a stereo
camera. The measurement process is carried out
in real-time, automatic, and documented. In this
study, we use the template matching method.
This method is an improvement in terms of pro-
cessing speed upon the existing correlation meth-
ods [15, 16]. The advantages of template match-
ing compared to background subtraction and
optical flow methods are in terms of the detec-
tion process, in which the position of the object is
2 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 0 6
 Development of laboratory devices for real-time measurement
independent of the previous frame. Through tem-
plate matching and stereo camera triangulation,
the measured object is selected and isolated dir-
ectly from the image produced by the stereo cam-
era so that the 3D object position is obtained.
Our motivation to do this work is to develop
a simple stereo camera system in the near
future, that can measure real-time 3D position
for educational purposes. Basic physics education
requires understanding fundamental physics the-
ories primarily related to the object motion in both
scalar and vector quantities. Examples of motion
that can will be reviewed include examples such as
bullet motion, pendulum oscillation motion, and
object motion on the slope. We expect this meas-
urement system can bridge the theories of physics
to its application in real situations related to the
scalar and vector presentations. This method of
measurement can also contribute to the proof of
theory through experimentation. Through experi-
ence using this measurement system, abstract the-
ories of physics can be applied in real-time so
that the process of learning physics is more mean-
ingful, fun and can improve analytical thinking
skills.
1.2. Theory and calculation
The theoretical approach is necessary to apply the
calculations used. There are two main concepts
used in this paper, namely coordinate transforma-
tion and triangulation model.
1.2.1. Image coordinate transformation. The
coordinates of the image are generally located at
the top left end. The transformation of coordinates
from the top left end (P) to the centre of the image
(O) is illustrated in figure 1.
Based on figure 1, an image with a resolution
of W × H with a horizontal coordinate axis (u)
to the right and a vertical (v) axis down can be
transformed into the center of the image using:
x = u − 0.5W (1)
y = 0.5 H − v. (2)
Variable x is the horizontal axis, and y is the
vertical axis defined from the center of the image.
March 2022
Figure 1. Image coordinate transformation from point
P to O.
1.2.2 Triangulation model. The model of the 3D
visual measurement method using a stereo cam-
era is shown in figure 2. In this method, the ref-
erence frame must be defined in advance. In this
case, the position of the left camera is selected as
the reference point for the coordinates of the real
space O(X, Y, Z). The position of the right cam-
era to the left camera is separated by b. A stereo
image plane is a plane where the stereo image is
formed. Physically the stereo image plane is the
same plane with both camera sensors. The dis-
tance of the stereo image plane depends to the
focus of both lenses. Since the focus of both lenses
is same, the stereo images formed exactly in the
same plane. However, there is a horizontal shift
of the right image to the left image. The object
distance in real space to the left lens is ZObject ,
whereas the reference points in the image pro-
duced by the left and right cameras are ol and or ,
respectively. The image of the object formed in
the left image and the right image will have a dif-
ferent position to the reference point. The differ-
ence in position against the reference is called the
disparity of d. Based on these definition the dis-
parity can be stated by:
d = (xr − xl ) , (3)
where xl and xr are the positions of objects in the
left and right images.
The relationship d with Z object can be stated
by:
fb
d= . (4)
Zobject
The values f and b depend on the specifications of
the stereo camera, thus on a fixed Z object , the value
3 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 0 6
S Ristanto et al
Figure 2. A model of the 3D visual measurement
method using a stereo camera.
of disparity between stereo cameras with each
other can be different. In stereo cameras where f
and b remain, the change in disparity depends only
on Z object changes.
Furthermore, assuming the distortion effect
is negligible, the relationship between the object
position in real space (X, Y, Z) and their position
in the stereo image (x, y) can be expressed by:
However, the records of the object position
on the image also depend on many factors. So,
the calibration process is still necessary to find
the proper relationship between the Z and d val-
ues, the Y and y values, and the X and x values.
Many reviews on the camera calibration process
are available extensively [22–26].
Zang et al [22] have proposed a self-
calibration method using each beam captured
by the camera as a basic adjustment unit. This
method works well when applied to Chang’e
cruisers on the Moon. Gai et al [23] has pro-
posed a two-camera calibration method for 3D
optical-based measurements. This method uses a
planar calibration plate that changes its orienta-
tion in the measurement area. Calibration para-
meters are obtained through imagery, and then
translation and rotation matrices are calculated
using a centroid distance increment matrix. Jia
et al [24] has proposed calibration of perpendic-
ularity compensation-based stereo cameras. This
method uses five images that have obtained preci-
sion well. Cui et al [25] proposed calibrating ste-
reo cameras by minimizing metric distance errors
between triangulation principles. Ma [26] has pro-
posed calibrating stereo cameras using 1D objects
that are free to move and then applying Heteros-
cedastic error in variables models to improve their
accuracy.

fb
Z= (5) 2. Materials and methods
d
The device comprises a built-in stereo camera,
bx tripod, and a mini-PC. The specifications of the
X= (6) stereo camera and the mini-PC are presented in
d
table 1.
The stereo camera was connected through
by
Y= . (7) the USB 3.0 port of the mini-PC. The software
d was developed in Python 3.7, and libraries of
To predict the accuracy of the measurement of OpenCV 4.0 equipped it [27]. The Python code
the distance of the object to the camera (Z) we can be found as an additional file (available online
propose the magnitude of resolution of disparity at stacks.iop.org/PED/57/025006/mmedia) of this
(RoD). RoD is defined as the ratio of distance to paper.
the disparity: Calibration and testing tools use kalibrator
board, water pass, and slider. The system setup is
Zmeasured shown in figure 3. The calibrator object was an
RoD = , (8) acrylic board with a grid pattern of 2 cm and con-
d
centric circles of a multiplied radius of 2.5 cm. We
where Z measured is the measured distance of the cal- also put a piece of isolation tape as a test object.
ibration board to the camera, and d is disparity. The test object could be moved easily around the

March 2022 4 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 0 6
 Development of laboratory devices for real-time measurement
Table 1. Camera and mini-PC specifications. 2.1.1. Stereo image capture. At this stage, the
Stereo camera Specifications step is to initialize the stereo camera. Then, set
the resolution and frame rate of the stereo image
Model Stereo 3D VR USB 2.0 according to its specifications. In this paper, the
Manufacturer www.kayetoncctv.com/ resolution of the image used is 1280 × 460 pixels.
Focus 3,0 mm (fixed) The resolution results from a combination of the
Sensor ¼ inch left and right cameras with 640 × 460 pixels res-
Pixel size 3 µm × 3 µm
olution. The left camera image is located in the
Object distance 30 cm—infinity
FoV 100◦
horizontal range (0–640) pixels, while the right
Image format MJPEG 2560 × 720 30 FPS camera image is in the range (641–1280) pixels. If
Baseline 6 cm this stage is successful, then on the monitor screen
appears a stereo image display.
Mini PC Specifications

Processor Intel(R)Core(TM) i5-8250U 2.1.2. Template definition. Before the measure-

CPU@1.6 GHz 1.8 GHz ment process, a template should be available. A
RAM 16 GB template is part of an image selected as a rep-
System type 64-bit OS, ×64-based Proc
resentative of an object in a real space. Likewise,
Operating system Windows 10
Software Python 3.7 and OpenCV 4.0
object positioning in real space is represented by
template displacement in the image.
Template selection can be made on both left
and right image. However, the template selec-
tion needs to be noted because the object to be
measured is in the field of view of both cameras
so that the object appears in the left and right
imagery.
In this paper, the template determination pro-
cess is determined using the mouse. The cursor is
directed at the object to be measured and pressed
left click then simultaneously produced a template
size 24 × 24 pixels. The template is automatically
saved in ‘JPG’ format. Templates can be changed
at any time, either before taking measurements or
Figure 3. Devices arrangement comprises a mini PC, during the measurement process.
a stereo camera, a slider and a calibration board with a
test object, and a water pass.
2.1.3. Template matching. Template matching
is matching templates on images produced by the
calibrator board, while the calibrator board could left camera and the right camera. The matching
be moved on a slider approaching the stereo cam- template procedure starts by calling the template
era. A slider was used to let the calibrator board file that has been generated in stage 2. Next, sep-
moving perpendicular to the image plane so that arate the left image and the right image on the
the calibration results were more accurate [24]. A resulting stereo image. The resolution of the left
water pass is used to make sure the camera plane image and the right image are 640 × 460 pixels,
is parallel to the calibrator board. respectively. Furthermore, a template resolution
of 24 × 24 pixels is sought after in the left image
area and the right image. The matching process
2.1. Method for object detection on stereo on both images is done separately simultaneously
image so that the matching process on the left and right
Stages in the process of measuring 3D positions image runs independently and does not interfere
are as follows: with each other. Areas that match the template are

March 2022 5 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 0 6
S Ristanto et al
marked with pink squares according to the tem-
plate size.
The template matching success indicator is
characterized by two things. Firstly, the appear-
ance of pink squares on the left and right image.
Secondly, the vertical location (y) of the box on
the left and right images are the same, while the
horizontal location (x) is different. Based on ste-
reo cameras, the height of the left camera and the
right camera are the same. The two cameras are
only separated horizontally. If both indicators are
met, then it can be continued to the following pro-
cess. If not, the meter can reselect the template
using the mouse as in stage 2. This process is done
manually by the gauge.
The next step determines the object location
in the template image on the left and right images.
The assumptions used by moving objects are fol- Figure 4. Position measurement algorithm using a
lowed by the homogeneous displacement of pixels template matching technique on stereoscopic images.
in the template. Assuming that, the object can be
represented the location of a pixel point on the 2.1.6. Result presentation. Once all measure-
template. The pixel point selected is the top left ment parameters have been successfully obtained,
end. the next step is to display the measurement res-
ults on the monitor screen. The magnitude shown
includes the position of the object in the image
2.1.4. Disparity calculation. The disparity is the (x, y) in pixels, the disparity (d) in pixels, and the
difference in the horizontal position of objects in 3D position (X, Y, Z) in cm.
the left image and the right image. The process of This stage also displayed the coordinate axis
calculating disparity starts from equation (3). The in the image. The center of the left image is selec-
xl and xr values are taken from the horizontal pos- ted as the reference point.
ition of the matching template on the left image Stages 1 to 6 are procedures in one meas-
and the right image. urement loop. Once the measurement process is
The validity of disparity in the test uses cri- complete, the results can be saved in videos in
teria d > 0 and the value yr = yl with d as dis- ‘AVI’ format and spreadsheets in ‘xlxs’ format.
parity, yr as the vertical position of objects on the The measurement algorithm is shown in figure 4.
right image, yl as the vertical position of objects
on the left camera. If it does not meet, then the
information ‘object not detected. Please, select 2.2. Method of calibration
the template again’. At this stage, the meter can The device for the calibration process was a rect-
change the template directly on the stereo image angular object being observed and a measure-
displayed on the monitor screen until the inform- ment system that used a stereo camera. The test
ation no longer appears. If valid, then continue in object was an isolation tape affixed to the calib-
the following process. rator board.
In this study, each coordinate component
is calibrated separately. The calibration process
2.1.5. Object position calculation. Once the dis- starts with the Z-axis component. This calibration
parity value is valid, the next step is to calcu- is done because it is based on equation (5).
late the 3D position of the object using equations The calibration process of the object dis-
obtained through calibration. The calibration pro- tance to the camera (Z) was done by changing
cess is described in section 3.2. the camera position gradually at 1 cm intervals on

March 2022 6 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 0 6
 Development of laboratory devices for real-time measurement
Figure 5. Stereo image display. The test object is in a pink box, as seen by the left camera and right camera.
the Z-axis, while X = Y = 0. The object distance
was changed in the range of 30 cm to 70 cm. The
closest distance of the object to the stereo cam-
era was 30 cm because it was the minimum dis-
tance that the object still in focus. Meanwhile, the
farthest distance was 70 cm because there is no
significant change in disparity at that distance and
after.
The next step is to calibrate the X-axis com-
ponents. The calibration process is done by gradu-
ally changing the object position every 2 cm
according to the size of the calibrator board box,
while Z = 40 cm, Y = 0. Objects are measured
from X = −9 cm to X = 9 cm according to the
range of stereo cameras. This process generates
the horizontal position relationship of the object
in real space (X) as a function of the horizontal
position of the object in the image (x), as stated in
equation (6).
Finally, calibrate the Y-axis components.
Objects are placed at Z = 40 cm, X = 0, while
Y is varied gradually every 2 cm from Y = −6 cm
to Y = 8 cm according to the field of view of ste-
reo cameras. Thus, the vertical object position in
real space (Y) as a function of the vertical object
position in the image (y) is obtained according to
equation (7).
2.3. Method position measuremnet test
Measurements are performed by varying the posi-
tion at (X, Y, Z) directions of the tested object. The
March 2022
measurement aims to determine the error propaga-
tion in each direction.
The measurement range was determined after
calibration data sets were obtained. Based on the
calibration data, the measurement range at certain
precision was obtained. In this study, a maximum
error of 1 cm was set, and so the measurement
range that meets the criteria was concluded.
2.4. Method for real-time performance
A computer timer facility is put at the beginning
and the end of the measurement process loop, as
described from stage-1 to stage-6. The measure-
ment time delay is calculated based on the dif-
ference between the end time and the initial time.
One measurement loop works to acquire a single
frame of the stereo image. Technically, if the cam-
era frame rate is set at 30 FPS, then the one-time
processing time of the measurement should be less
than or equal to 33.3 ms. If the processing time
is greater than 33.3, even if the frame rate is set
at 30 FPS, the number of frames produced per
second is less than 30.
Tests were carried out to see the influence
of the computers that were used in real-time
measurements. The real-time measurement by a
mini PC with an Intel processor of Core i5 and
RAM 16 Gb and a notebook with an Intel Pro-
cessor of Core i3 and RAM 4 Gb were compared.
The number of total measurements was 100 times.
7 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 0 6
S Ristanto et al

Figure 6. The disparity profile represents the effect of Figure 7. Change in RoD along the Z-axis.
the object to camera distance to the disparity.

3. Result and discussion

3.1. Object detection on stereo image
Automatic and real-time object detection results
in stereo images are shown in figure 5. The right
camera and left camera images are combined into
one. The detection results of template matching
are marked as a pink rectangle on the left and right
camera images. The measurement results are dis-
played at the bottom of the image. The measure-
ment results can also be displayed automatically
and in real-time. The images were stored in an
AVI format video, and the data were saved as a
spreadsheet in MS Excell document. The quant-
ities stored in the spreadsheet include disparities,
the position of objects in the image (x, y), and the
positions of objects in space (X, Y, Z) as shown on
the stereo image. The disparity (d) was calculated
using equation (3).

3.2. Calibration process

The graph of the calibration results on the Z-axis
is shown in figure 6. The graph shows that the rela-
tionship between the object distance to the camera
(Z) and the disparity is non-linear, in which the
farther the object being measured, the smaller the
disparity. It is relationship is presented by: Figure 8. The results of the measurement calibration
for the X-axis (a) and the Y-axis (b).
d = 4039.8Z−0.999 . (9)

Equation (9) is following equation (4) with a cor- and the baseline b in cm. Then taking the inverse
relation value of 0.9997. The value of 4039.8 is of the equation (9) will provide the distance Z is a
the multiplication constant of lens focus f in pixels function of disparities.

March 2022 8 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 0 6
 Development of laboratory devices for real-time measurement
Table 2. The results of the measurement of the object’s position.

Object position (cm) Position measurement (cm)

No. X0 Y0 Z0 X ± ∆X Y ± ∆Y ZX ± ∆Z

Z varied, X = Y = constant
1 3 0 30 2.9 ± 0.1 −0.2 ± 0.2 30.3 ± 0.3
2 3 0 35 2.9 ± 0.1 0.0 ± 0.0 35.3 ± 0.3
3 3 0 40 3.2 ± 0.2 1.0 ± 0.0 40.2 ± 0.2
4 3 0 45 3.3 ± 0.3 2.0 ± 0.0 45.7 ± 0.7
5 3 0 50 3.2 ± 0.2 3.0 ± 0.0 50.8 ± 0.8
6 3 0 55 3.6 ± 0.6 4.0 ± 0.0 55.7 ± 0.7
7 3 0 60 3.4 ± 0.4 5.0 ± 0.0 60.7 ± 0.7
X = varied, Z = Y = constant
1 −5 0 30 −5.0 ± 0.0 6.0 ± 0.0 29.7 ± 0.3
2 −3 0 30 −3.0 ± 0.0 7.0 ± 0.0 30.1 ± 0.1
3 −1 0 30 −1.1 ± 0.1 8.0 ± 0.0 30.3 ± 0.3
4 1 0 30 0.9 ± 0.1 9.0 ± 0.0 30.5 ± 0.5
5 3 0 30 2.8 ± 0.2 10.0 ± 0.0 30.1 ± 0.1
6 5 0 30 4.7 ± 0.3 11.0 ± 0.0 29.7 ± 0.3
Y = varied, Z = X = constant
1 0 6 30 0.1 ± 0.1 6.0 ± 0.0 29.9 ± 0.1
2 0 4 30 0.0 ± 0.0 4.0 ± 0.0 30.1 ± 0.1
3 0 2 30 0.1 ± 0.1 1.9 ± 0.1 29.9 ± 0.1
4 0 0 30 0.0 ± 0.0 0.0 ± 0.0 29.9 ± 0.1
5 0 −2 30 0.0 ± 0.0 −2.0 ± 0.0 30.1 ± 0.1
6 0 −4 30 0.1 ± 0.1 −3.9 ± 0.1 30.1 ± 0.1
7 0 −6 30 0.0 ± 0.0 −5.9 ± 0.1 29.9 ± 0.1

The graph pattern produced in this study is
similar to the previous study results [1], with an
improvement in the graph correlation coefficient
of R2 = 0.9997. Figure 6 is also identical to the ste-
reo camera error modeling done previously [28].
The non-linear change in disparity causes the
RoD also change along the Z-axis.
The change in RoD on the Z-axis is shown in
figure 7.
Figure 7 shows that the RoD increases by
quadratic magnification as the object position gets
further from the camera. The graph can determine
the accuracy of the measurement at a specific posi-
tion from the camera. For example, if the expected
accuracy of the measurement results in less than
1 cm, then the recommended distance of the object
should be in the range of 30–60 cm. This range is
chosen by the error assumption of 1 pixel so that
a measurement error varies in the range of 0.2–
0.9 cm.
The X-axis and Y-axis calibrations are shown
in figure 8. Figure 8(a) shows that the relationship
of the X values in real space to x values in the
images as expressed in equation (6) is approxim-
ated linearly with a correlation of 0.9997. Like-
wise, figure 8(b) shows the relationship of the
Y values to the y values in the image following
equation (7) with a correlation of 0.9999. There-
fore, the X and Y can be calculated by equations
(6) and (7).
3.3. Position measurement test result
After calibration, the next step is to test the meas-
urement result. The object being measured is
placed in nine different places and then its position
is measured. Data on the measurement trial pro-
cess are presented in table 2. The 3D position of
the object in real space is expressed in R0 (X 0 , Y 0 ,
Z 0 ) while the measurement results are expressed
in R (X, Y, Z). Errors for each axis are expressed
in terms of ∆X, ∆Y, ∆Z.
Based on table 2, the measurements of sev-
eral 3D positions of the object can be shown.

March 2022 9 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 0 6
S Ristanto et al
Figure 9. Distribution of time delay of 100
measurements.
The recorded values of the positions of the X, Y,
and Z components match the 3D position of the
object being measured. The maximum error for
each axis is 0.6 cm for ∆X, 0.2 cm for ∆Y, and
0.8 cm for ∆Z.
This accuracy is still different from the
accuracy of the previous distance measurement
between two positions of 0.5 mm [5]. This accur-
acy is due to differences in the measurement
range and the methods used. Meanwhile, 3D posi-
tion measurement using Opto-Acoustics yields an
accuracy of 2.2 cm [29] in which has no signific-
ant difference from the results in this study.
3.4. Real-time performance
The result of the time delay measurement is shown
on the time delay graph as a function of the meas-
urement sequence, as shown in figure 9.
The average processing time depended on
the computer specification used. Using Processor
Intel Core i3 and i5, both provided different time
processing of (52 ± 3) ms and (32 ± 15) ms,
respectively. However, these results were faster
than the optical flow method at 220 ms [15] and
162 ms [16]. An interesting error contribution
when using the Processor Intel Core i5 seemed to
be due to inevitable internal processing. Based on
these test results, the Intel Core i3 processor pro-
duces a maximum frame rate of 19 FPS, while the
Intel Core i5 processor can produce a maximum
frame rate of 30 FPS.
March 2022
4. Conclusions
This research has developed the device and
method for measuring the 3D position using
a stereo camera that automatically measures a
moving object’s position in real-time and is well
documented. The method uses template matching
and triangulation techniques. The calibration res-
ults show that the object position in the z-direction
depends on the disparity, while the position of the
object in the direction of x and y remains constant.
The maximum accuracy of measuring the 3D pos-
ition in the X-axis is 0.6 cm, the Y-axis is 0.2 cm,
and the Z-axis is 0.8 cm in range X(−5–5) cm,
Y(−6–6) cm, and Z(30–60) cm, respectively. The
average processing time depends on the computer
specification being used. However, both Processor
Intel Core i3 and i5 provide timely processing of
less than 50 ms, which is regarded as a real-time
process.
Data availability statement
All data that support the findings of this study are
included within the article (and any supplement-
ary files).
Acknowledgments
We express our gratitude to the Ministry of
Research, Technology, and Higher Educa-
tion/Kemdikbud for providing BPPDN Grants and
the 2020–2021 Doctoral Dissertation Program so
that this research can be carried out. Thank you
to PT Madeena Karya Indonesia for preparing
the stereo camera-based measuring instrument
prototype used in this research.
ORCID iDs
Sigit Ristanto  https://orcid.org/0000-0003-
1395-8739
Gede B Suparta  https://orcid.org/0000-0002-
9048-5560
Received 16 September 2021, in final form 8 November 2021
Accepted for publication 2 December 2021
https://doi.org/10.1088/1361-6552/ac3f72
10 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 0 6
 Development of laboratory devices for real-time measurement
References
[1] Liu Z and Chen T 2009 Distance Measurement
System Based on Binocular Stereo Vision
Int. Jt. Conf. Artif. Intell. 456–9
[2] Sun X, Jiang Y, Ji Y, Fu W, Yan S, Chen Q, Yu B
and Gan X 2019 IOP Conf. Ser.: Earth
Environ. Sci. 252 052051
[3] Mahammed M A, Melhum A I and Kochery F A
2013 Object Distance Measurement by
Stereo Vision Int. J. Sci. Appl. Inf. Technol.
2 5–8
[4] He W, Wang J and Fu Y 2018 Creepage Distance
Measurement Using Binocular Stereo Vision
on Hot-line for High Voltage Insulator Curr.
Opt. Photon. 2 348–55
[5] Wang X, Liu J, Liu S, Jin P, Wu T and Wang Z
2018 Measurement 117 326–38
[6] Barone S, Neri P, Paoli A and Viviano A 2018
Proc. Struct. Integr. 12 122–9
[7] Rodríguez-Quiñonez J C, Sergiyenko O,
Flores-Fuentes W, Rivas-Lopez M,
Hernandez-Balbuena D, Rascón R and
Mercorelli P 2017 Opto-Electron. Rev.
25 24–32
[8] Yang T W, Zhu K, Ruan Q Q and Han J D 2008
Moving Target Tracking and Measurement
with a Binocular Vision System 15th
International Conference on Mechatronics
and Machine Vision in Practice 15th Int.
Conf. on Mechatronics and Machine Vision
in Practice (Auckland, New Zealand,
2–4 December 2008) pp 85–91
[9] Murmu N, Chakraborty B and Nandi D 2019
Measurement 141 1–11
[10] Sharma R D, Agrwal S L, Gupta S K and
Prajapati A 2017 Optimized Dynamic
Background Subtraction Technique for
Moving Object Detection and Tracking 2nd
Int. Conf. on Telecommunication and
Networks (TEL-NET) 2nd Int. Conf. on
Telecommunication and Networks
(TEL-NET) (Noida, India, 10–11 August
2017) pp 1–3
[11] Srivastav N, Agrwal S L, Gupta S K,
Srivastava S R, Chacko B and Sharma H
2017 Hybrid Object Detection using
Improved Three Frame Differencing and
Background Subtraction 7th International
Conference on Cloud Computing, Data
Science & Engineering - Confluence 7th Int
Conf Cloud Comput Data Sci Eng—Conflu
(Noida, India, 12–13 January 2017)
pp 613–7
[12] Babaee M, Tung D and Rigoll G 2018 A deep
convolutional neural network for video
sequence background subtraction Pattern
Recognit. 76 635–49
March 2022
[13] Szeliski R 2021 Computer Vision Algoritma and
Applications 2nd edn (Berlin: Springer)
[14] Kienholz C, Amundson J M, Motyka R J,
Jackson R H, Mickett J B, Sutherland D A,
Nash J D, Winters D S, Dryer W P and
Truffer M 2019 Tracking icebergs with
time-lapse photography and sparse optical
flow, LeConte Bay, Alaska, 2016–2017
Journal of Glaciology 65 195–211
[15] Su H, Chen Y, Tong S and Zhao D 2019 9th Int.
Conf. on Information Science and
Technology (ICIST) pp 350–6
[16] Guan W, Chen X, Huang M, Liu Z, Wu Y and
Chen Y 2018 IEEE Photon. J. 10 1–22
[17] Queirós S, Vilaça J L, Morais P, Fonseca J C,
D’hooge J and Barbosa D 2017 Int. J.
Numer. Method Biomed. Eng. 33
[18] Yang H, Huang C, Wang F, Song K and Yin Z
2019 IEEE Trans. Image Process.
28 3061–74
[19] Singh S, Khosla A and Kapoor R 2019 Visual
Object Tracking by Fusion of Audio
Imaging in Template Matching Framework
Int. J. Image Graph Signal Process. 8 40–49
[20] Liu F, Gong C and Huang X 2018 IEEE Trans.
Image Process. 27 2777–90
[21] Chen Z 2019 IEEE Access 7 88870–80
[22] Zhang S, Liu S, Ma Y, Qi C, Ma H and Yang H
2017 ISPRS J. Photogramm. Remote Sens.
128 287–97
[23] Gai S, Da F and Dai X 2018 A novel dual-camera
calibration method for 3D optical
measurement Opt. Lasers Eng. 104 126–134
[24] Jia Z, Yang J, Liu W, Wang F, Liu Y, Wang L,
Fan C and Zhao K 2015 Improved camera
calibration method based on
perpendicularity compensation for binocular
stereo vision measurement system Opt.
Express 23 15205–23
[25] Cui Y, Zhou F, Wang Y, Liu L and Gao H 2014
Opt. Express 22 9134–49
[26] Ma H 2017 A new calibration method for
binocular camera based on fundamental
matrix & HEIV model Adv. Model. Anal. A
54 333–45
[27] Ristanto S, Nugroho W, Sulistya E and
Suparta G B 2021 System and Method for
Stereoscopic Image Acquisition AIP Conf.
Proc. 2374 020014-1–7
[28] Gross E 2015 Physica A 430 1–10
[29] Esslinger D, Rapp P, Wiertz S, Sawodny O and
Tarin C 2018 Highly Accurate 3D Pose
Estimation for Economical Opto-Acoustic
Indoor Localization 15th International
Conference on Control, Automation,
Robotics and Vision (ICARCV) (Singapore,
18–21 November 2018) 1984–90
11 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 0 6
S Ristanto et al
Sigit Ristanto, M.Sc. is a PhD
candidature at the Department
of Physics, Universitas Gadjah
Mada, Indonesia. He also works
as a lecturer at the Department of
Physics Education, Universitas PGRI
Semarang, Indonesia. He mainly
works on the Physics of Imaging and
uses that for developing experimental
tools based on real situations rather
than laboratory bench that are useful
in physics education.
Waskito Nugroho, Ph.D. (RIP).
Worked as a lecturer at the Depart-
ment of Physics, Universitas Gadjah
Mada, Indonesia. Concerns on
programming and measurement
based on Physics of Imaging. Some
works related with Non-Destructive
Testing, Imaging Physics, Applied
Nuclear Physics, and Laser-Induced
Breakdown Spectroscopy (LIBS).
March 2022
Eko Sulistya, Ph.D. Worked as a
lecturer at the Department of Physics,
Universitas Gadjah Mada, Indonesia.
Mainly works on Computational
Physics and developing educational
kit using sensor and Arduino. Also
is the Head of the Basic Physics
Laboratory, in charge of carrying
out basic physics experimental
services and developing experimental
equipment and methods.
Prof. Gede Bayu Suparta, Ph.D.,
works as a full Professor at the
Department of Physics, Universitas
Gadjah Mada, Indonesia. Mainly
works on Physics of Imaging for
visual inspection, industrial imaging,
laboratory imaging, medical imaging,
digital radiography and computed
tomography.
12 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 0 6
 PAPER
Phys. Educ. 57 (2022) 025024 (10pp) iopscience.org/ped

Comprehensive
observations on pendulum
oscillation using stereo
vision
Sigit Ristanto1,2, Waskito Nugroho1, Eko Sulistya1
and Gede B Suparta1,∗
1
Department of Physics, Universitas Gadjah Mada, D.I Yogyakarta 55281, Indonesia
2
Departement of Physics Education, Universitas PGRI Semarang, Semarang 50232,
Indonesia

E-mail: gbsuparta@ugm.ac.id

Abstract
We have developed a stereo vision system for observing real objects for
experimental physics demonstration. The stereo vision observes objects from
two different points of view. An observer can identify the distance and
velocity of an object based on at least two different points of view. This
paper demonstrates our comprehensive observation in a pendulum oscillation
experiment to test the stereo vision system where a stereo camera is in a
forward view position. The stereo camera produces stereo images. We use
automatic object tracking techniques to obtain the 3D position of the
pendulum in real space based on the sequential stereo images of the
pendulum movement. The experimental results in a 3D graph that shows the
trajectory of the pendulum oscillation. We confirm the common
understanding of the oscillation movement by obtaining the graph of the
pendulum oscillation as a function of time. We can obtain the oscillation
motion parameters such as amplitude, angular frequency, frequency, period,
and initial phase. Also, we can show the damping oscillation phenomena and
the damping factor.

Keywords: pendulum experiment, stereo vision, damping oscillation

Supplementary material for this article is available online

∗
Author to whom any correspondence should be addressed.

1361-6552/22/025024+10$33.00 1 © 2022 IOP Publishing Ltd

S Ristanto et al
1. Introduction
The pendulum was known early in the develop-
ment of science. Galileo Galilei first observed
the pendulum motion as an early scientist in the
development of science [1]. Several decades later
followed the use of a pendulum to measure the
gravity acceleration of the earth. At that time, ana-
logue clocks also used the pendulum. Next, scient-
ists used a pendulum to prove the existence of the
Coriolis force due to the Earth rotation.
The pendulum has a typical trajectory. The
motion of the pendulum is affected by the force
of gravity. The pattern of motion identifies the
role of other forces. Also, many familiar pen-
dulum motion parameters include angular fre-
quency, amplitude, frequency, period, and initial
phase.
As with the analysis of motion in general,
there are two approaches in mechanics for analyz-
ing the pendulum, namely the force approach and
the energy approach. The force approach shows
that the driving force makes the oscillating motion
of the pendulum, which is a force proportional to
the object’s position, which always leads to the
equilibrium point. On the other hand, the energy
approach shows that if the pendulum system is
isolated so that the total energy of the pendulum is
always constant, the oscillation motion of the pen-
dulum is a constant change from potential energy
to kinetic energy and vice versa. If there is addi-
tional energy from the environment, the amplitude
will be even more significant. Otherwise, if there
is energy wasted in the environment, the oscil-
lation amplitude will decrease. However, both
approaches produce the same oscillation equation.
The pendulum has been applied in various
fields as follows. The inverted pendulum system
for robotic and manufacturing applications may
apply a fuzzy control [2, 3]. The piezoelectric
efficiency in the energy harvesting process may
increase using the double pendulum concept [4],
while another energy harvesting may also use the
concept of an auto-parametric pendulum absorber
[5]. The pendulum concept has proven to dampen
vibrations experienced by offshore wind power
plants [6, 7]. Efforts to refocus high-rise structures
using the pendulum concept have proven to work
well [8].
March 2022
Pendulum motion analysis is necessary to
dynamically stabilize and control feedback on
the pendulum according to the desired position
[9]. Theoretical solutions when the length of
the string of the pendulum changes shows inter-
esting motion. When the length of the string
changes quadratic with time, the amplitude is
constant. However, when the length of the rope
changes sinusoidally, an oscillation explosion
effect occurs [10]. Measurement of the pendu-
lum parameters such as amplitude, frequency,
etc is important in the pendulum applications
as described above. The stereo vision is one of
the measurement methods that can be applied to
measure the parameters of the pendulum oscilla-
tion in 3D space.
The development of stereo vision began sev-
eral decades ago. In 1979 the stereo matching
algorithm was introduced [11]. Stereo matching is
matching the image similarity on the left and right
of the stereo image. In the last 5 years, the stereo
vision has overgrown. Some reviews on the influ-
ential parameters have increased the stereo vis-
ion accuracy [12]. The accuracy of 3D distance
measurement using stereo vision can be improved
using a neural network [13]. Moreover, a robotic
arm system for systematic calibration based on
stereo vision has been successfully developed
[14]. A calibration method to improve accuracy
in 3D location determination has been developed
[15]. A review of the stereo image acquisition sys-
tems and methods can be found in [16].
Applications of stereo vision have been in
various fields, such as in the transportation sec-
tor, road inspection systems [17], early warnings
before collisions [18], and automatic car steer-
ing [19]. In the fields of photogrammetry, out-
door mapping [20], indoor navigation [21], and
underwater cave mapping [22] use stereo vision.
However, references to the use of stereo vision for
observing pendulum motion are still limited.
One of the studies directly related to this topic
is stereo vision to detect the trajectory of thrown
objects. The trajectory of a sheeted object is also
affected by gravity, as is the oscillation of a pen-
dulum. These studies can track the object traject-
ory with an accuracy of 1 cm in 85% of cases
and 3 cm in 92% of cases [23]. Another related
2 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 2 4
 Comprehensive observations on pendulum oscillation using stereo vision
research is using genetic algorithms to predict the
trajectory of thrown objects using stereo vision.
The simulation results using a numerical approach
show that the method can detect object trajectories
accurately [24].
Several studies have tried to measure pen-
dulum oscillation parameters. Each study focuses
on a specific parameter, such as measurement of
frequency and damping coefficients using sensors
on computer mouse hardware [25], period meas-
urements using photogate sensors [26], amplitude
and period using wireless centripetal acceleration
measurements [27], pendulum trajectory using a
camera [28], and measurement of pendulum oscil-
lation damping coefficient using resonance phe-
nomena [29].
This paper proposes a method to observe all
aspects of the motion of the pendulum oscilla-
tion from the forward view using stereo vision.
Stereo vision plays observation of objects from
two different sides, similar to how the two human
eyes work. The two human eyes together can
estimate the distance of an object to produce dif-
ferent points of view. In this paper, we present
our comprehensive observations on the pendu-
lum and its trajectory, the graph of the pendulum
oscillation as a function of time, and the para-
meters of the oscillation motion such as amp-
litude, angular frequency, frequency, period, and
initial phase. We aim to show and confirm such
phenomenon experimentally using the method
developed.
Observation on the motion of the pendulum
uses a stereo camera to produce a stereo image.
The object tracking method can be applied to
the stereo image to obtain the disparity of the
pendulum position in the image. Then, based on
the disparity information, the 3D position of the
pendulum can be obtained through the triangu-
lation method. After obtaining the 3D position,
observations continue from the forward-looking
side.
We expect to extend the application of this
method for the observation of various oscilla-
tion motions such as mass-spring oscillations, 3D
oscillations, Foucault pendulum oscillations, and
torsion pendulums.
March 2022
2. Method
2.1. Materials
The equipment used in this research is a set of
stereo cameras and a mini PC, along with the
software capable of measuring 3D position. The
software performance is well calibrated.
The description of the stereo camera specific-
ations is as follows. The distance between the left
and the right cameras (b) is 6 cm. The camera
sensor measures 0.25 inches. The image resolu-
tion of 1280 × 480 pixels is a combination of the
right and left camera images, and each has a resol-
ution of 640 × 480 pixels, respectively. The field
of view of the camera is 100◦ . Observation is for
the distance of objects in the range of 30 cm to
infinity. The camera frame rate is 30 FPS.
The description of mini PC specifications is
as follows. The processor used is an Intel Core
i5 with 16 GB of RAM. System type 64 bit. The
operating system uses Windows 10. We developed
the stereo vision software using the Python pro-
gramming language, equipped with the OpenCV
library.
2.2. Theory and calculation
The pendulum oscillation observed in this study
is the change in the object distance to the stereo
camera. Measurement of the object distance to the
camera in stereo vision is using the triangulation
method. An illustration of stereo vision is shown
in figure 1.
Based on figure 1 we conclude
b = (Xr + Xl ) (1)
xr
Xl = Z (2)
f
−xl
Xl = Z . (3)
f
By inserting equations (2) and (3) into equation
(1), we obtain equation (4)
Z Z
b= (xr − xr ) = d (4)
f f
3 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 2 4
S Ristanto et al
Figure 1. The concept of stereo vision triangulation.
The horizontal distance of the pendulum on the left
image is xl while the right image is xr . The focus of
the lenses of both cameras is the same, namely f. The
distance between the two cameras b. The distance of
the pendulum to the camera Z. The horizontal distance
of the pendulum at the centre of the left camera is X l ,
while at the centre of the right camera is X r .
with the disparity d defined as
d ≡ (xr − xl ) . (5)
Disparity d is the difference between the hori-
zontal position of the pendulum on the right image
and the left image adjacently.
The centre of the left image coordinates is
regarded as the centre of the image coordinates as
well as the centre of the Cartesian coordinates of
real space. Therefore, equation (2) can be rewrit-
ten by equation (6)
x l
X=Z (6)
f where g is the acceleration due to gravity and l is
while equation (7) is valid in Y-axis
y
Y=Z . (7)
f A (t) = Ao e−∝t
Based on equations (4), (6), and (7), the rela-
tionship between the pendulum position in image
coordinates (x, y, z) and the pendulum position in
Cartesian coordinates (X, Y, Z) in real space can
be written as
fb
Z= (8)
d (3) α < ω Underdamping [30].
March 2022
bx
X= (9)
d
by
Y= . (10)
d
Furthermore, equations (8)–(10) are used as ref-
erences during the calibration process to obtain a
definitive relationship between the pendulum pos-
ition in the image coordinates and the pendulum
position in the Cartesian coordinates in real space.
Furthermore, the oscillating motion of the
pendulum on the Z-axis can be expressed by
∆Z = ∆Zmax sin (ωt + θ) (11)
where ∆Z max as the amplitude, ω as the angular
frequency, and θ as the initial phase.
Other oscillation motion parameters are fre-
quency and period. Frequency is the number of
vibrations in 1 s, while the period is the time it
takes the pendulum to make one vibration. The
relationship of frequency (v) and period (T) to the
angular acceleration can be expressed by
2π
ω= = 2πv (12)
T
where π is a constant of 3.14159. The angular fre-
quency (ω) concerning the parameters of the pen-
dulum motion can be expressed by
√
g
ω= (13)
the length of the pendulum string.
The damping of the oscillation amplitude A(t)
can be expressed by
(14)
where A0 is the initial amplitude, α is a damping
factor, and t is time. Based on the value of angu-
lar frequency and the damping factor, the damping
amplitude has three categories, namely
(1) α > ω Overdamping
(2) α > ω Critical damping
4 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 2 4
 Comprehensive observations on pendulum oscillation using stereo vision
Figure 2. Flowchart of stereo vision.
2.3. Flowchart of stereo vision
The flowchart of observing pendulum oscillations
using stereo vision is shown in figure 2.
The first step starts with recording a stereo
camera. At this stage, the camera generates a ste-
reo image with a resolution of 1280 × 480. The
second step is to detect the pendulum in the ste-
reo image. The third step is to apply stereo match-
ing to obtain the position of the pendulum on the
stereo. In the fourth step, the disparity is obtained
based on the difference in the pendulum posi-
tion on the stereo image. The fifth step is to cal-
culate the 3D position of the pendulum in real
space coordinates based on the calibration res-
ults. Step six, display a 3D graph of the traject-
ory of the pendulum. The seventh step displays
the oscillation of the pendulum and obtains all
its parameters. At this stage, the graph displays
the pendulum oscillation as a function of time.
A curve fitting was applied to find the appropri-
ate oscillation equation, and so all the oscillation
parameters.
March 2022
Figure 3. Experimental observation of pendulum
oscillations from the forward view based on stereo
vision.
2.4. Design of experiment
The experimental setting is shown in figure 3. A
pendulum is attached to a string 38 cm long. The
stereo vision system consists of a stereo camera
and a Mini PC to the left of the pendulum. The
initial distance of the pendulum from the camera
is Z = 37.5 cm. Then the pendulum is deviated by
6 cm in the direction of the stereo camera plane
so that the pendulum oscillation is in line with
the direction of the stereo camera plane. Further-
more, the stereo vision system observes the pen-
dulum oscillations. The change in the distance of
the pendulum from the camera (∆Z) characterizes
the pendulum oscillation.
3. Result and discussion
3.1. Image stereo
The data obtained is stereo video observations
in ‘AVI’ format, complemented by real-time and
automatic pendulum position measurements in 3D
(X, Y, Z). In addition, the spreadsheet records the
measurement data set of the 3D position and its
time. A snapshot image of the stereo vision sys-
tem from the forward-looking side is shown in
figure 4.
Figure 4 shows a stereo image consisting of
two images, namely the left and right images.
Each image has a white cross-line that is rep-
resenting the image centre. The centre point (0,
0) is at the centre of the left image. The images
show that the pendulum has the same vertical pos-
ition but is different in a horizontal position. This
phenomenon is called disparity, as described in
5 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 2 4
S Ristanto et al
Figure 4. A Snapshot image of the stereo vision system from the forward-looking side.
equation (5). A square shape located on the pendu-
lum characterizes successful object detection. The
caption in the above stereo image indicates the
disparity value and position of the pendulum, and
the caption at the bottom of the stereo image indic-
ates the 3D position of the pendulum. The exper-
iment found that the disparity value is 128 pixels,
while the pendulum position in the image is at
x = 23 pixels and y = 14 pixels. Therefore, we
can calculated the position of 3D pendulum in
real space as X = 1.1 cm; Y = 0.7 cm; and
Z = 31.8 cm. The developed software has incor-
porated all formulas.
The experimental data set obtained in this
experiment is a stereo video of the pendulum
oscillation accompanied by the 3D position meas-
urement. Students in the laboratory can take
advantage of the stereo video as they can see the
3D effect using e.g. a VR BOX. In addition, the
measurement data is stored in a spreadsheet so that
it can be recalled for further analysis. A sample of
stereo video is provided along with this paper.
3.2. Trajectory of pendulum oscillation
The results of measuring the 3D position of
the pendulum show the trajectory of the pendu-
lum oscillation. The trajectory of the pendulum
based on the measurement results is shown in
figures 5(a)–(c).
Figure 5(a) shows the trajectory of the pen-
dulum seen from a 3D perspective. Figure 5(b)
shows that the pendulum oscillations are on the Z-
axis with Z = 31.8 cm to Z = 42.9 cm. While the
March 2022
Y-axis is in the range 0 cm < Y < 1 cm. In addition,
it appears that the pendulum trajectory is parabolic
in the Z–Y plane. Figure 5(c) shows the trajectory
of the linear pendulum in the Z–X plane with a
correlation of 0.98. It also shows that the traject-
ory of the pendulum forms an angle as large as
tan−1 (1.1 – ((−1.1)/(42.9 – 31.8))) = 11◦ against
the Z-axis. Figure 5(d) shows that the pendulum
oscillations are on the X-axis with X = −1.2 cm
to X = 1.2 cm, while on the Y-axis is in the range
0 cm < Y < 1 cm. In addition, it shows that the
pendulum trajectory is parabolic in the X–Y plane
with a correlation of 0.94.
3.3. Pendulum oscillations
A sample graph of pendulum oscillations as a
function of time for 3 s is shown in figure 6.
Based on figure 6, it appears that for 3 s, the
pendulum has oscillated 2.5 s. The curve fitting
provides the oscillation parameters. The curve fit-
ting is arranged using equations (11) and found
that the correlation coefficient between the meas-
urement result and the curve fitting is 0.997. The
result of curve fitting is shown in equation (15),
( ( π ))
∆Z = 5.5 sin 5.094t − cm. (15)
2
Based on equations (15) and (12), the pendulum
oscillation motion parameters obtained are shown
in table 1.
The gravitational acceleration constant is
obtained based on the information obtained in
table 1 and equation (13). By entering the value
6 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 2 4
 Comprehensive observations on pendulum oscillation using stereo vision

Figure 5. Pendulum oscillation trajectory (a) 3D perspective view (b) display in the Z–Y plane (c) display in the
Z–X plane (d) display in the X–Y plane.

Table 1. Parameters obtained from the oscillating

motion of the pendulum.
No Parameters Value

1 Angular frequency 5.094 (rad s−1 )

2 Amplitude 5.5 (cm)
3 Frequency 0.81 (Hz)
4 Period 1.2334 (s)
5 Initial phase 0.5π

3.4. Damping phenomenon

Figure 6. Graph of pendulum oscillation as a function
of time.
of the angular frequency and the length of the rope
used of 38 cm into equation (13), the acceleration
due to gravity is 9.86 m s−2 . This result is almost
commensurate with the value of gravity using a
photogate sensor of 9.76 m s−2 [25] and a reson-
ance method of 9.74 m s−2 [28].
The damping phenomenon can be identified after
the pendulum has been oscillating for a long
time. In this study, the damping phenomenon was
identified after the pendulum oscillated for 175
seconds as shown in the figure 7.
The damping equation in figure 7 is expressed
as
A (t) = 5.5e−0.0028t (16)

March 2022 7 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 2 4
S Ristanto et al
Figure 7. The phenomenon of damping in pendulum oscillations.
where the initial amplitude is 5.5 cm, and the
damping factor is 2.8 × 10−3 s−1 . Accord-
ing to Fowles and Cassiday [30], this value
is much smaller than the angular frequency of
5.094 rad s−1 , so the damping is very weak. These
results are commensurate to Ladino [28] but with
a different approach.
We attempt to show how a stereo camera can
be used to record oscillation and then do proper
measurement based on the parameters related to
the stereo camera such as disparity and triangu-
lation. We have obtained some key parameters
related to the oscillation motion of the pendu-
lum, such as frequency, amplitude, period, and
initial phase. For further work, the influence of
the length of the pendulum, gravity, and the factor
of isochronism may be explored. This work may
be completed by doing graphical simulation or
numerical simulation. Yet, it is good to show
in a 3D picture when the energy of the pendu-
lum changes by its position. Therefore it needs
to measure the instantaneous height (Y-axis)
and the instantaneous velocity of the pendulum
simultaneously.
4. Conclusion
The results showed successful observation of the
phenomenon of pendulum oscillation. Compre-
hensive understandings provide the 3D graph
of the pendulum oscillation trajectory and the
pendulum oscillation graph. Also, it provides
the parameters of the oscillation motion such as
March 2022
amplitude, angular frequency, frequency, period,
initial phase, and damping factor. This obser-
vation method has contributed to expanding the
method of observing pendulum oscillations which
were initially only limited to the forward view dir-
ection. This method can also be applied to observe
other oscillation motions.
Data availability statement
All data that support the findings of this study are
included within the article (and any supplement-
ary files (available online at stacks.iop.org/PED/
57/025024/mmedia)).
Acknowledgments
All authors express the financial support for
this research from the Ministry of Research,
Technology, and Higher Education of Indone-
sia through BPPDN grants and the 2020–2021
Doctoral Research Grant, and from PT Madeena
Karya Indonesia for the prototype of the stereo
camera.
ORCID iD
Gede B Suparta  https://orcid.org/0000-0002-
9048-5560
Received 4 September 2021, in final form 10 December 2021
Accepted for publication 20 December 2021
https://doi.org/10.1088/1361-6552/ac44fc
8 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 2 4
 Comprehensive observations on pendulum oscillation using stereo vision
References
[1] Baker G L and Blackburn J A 2005 The
Pendulum: A Case Study in Physics (Oxford:
Oxford University Press)
[2] Su X, Xia F, Liu J and Wu L 2018
Event-triggered fuzzy control of nonlinear
systems with its application to inverted
pendulum systems Automatica
94 236–48
[3] Roose A I, Yahya S and Al-Rizzo H 2017
Fuzzy-logic control of an inverted
pendulum on a cart Comput. Electr. Eng.
61 3–47
[4] Izadgoshasb I, Lim Y Y, Tang L, Padilla R V,
Tang Z S and Sedighi M 2019 Improving
efficiency of piezoelectric based energy
harvesting from human motions using
double pendulum system Energy Convers.
Manage. 184 559–70
[5] Zhang A, Sorokin V and Li H 2021 Energy
harvesting using a novel autoparametric
pendulum absorber-harvester J. Sound Vib.
499 116014
[6] Jahangiri V, Sun C and Kong F 2021 Study on a
3D pounding pendulum TMD for mitigating
bi-directional vibration of offshore wind
turbines Eng. Struct. 241 112383
[7] Larsen T G, Zhang Z and Høgsberg J 2021
Vibration damping of an offshore wind
turbine by optimally calibrated pendulum
absorber with shunted electromagnetic
transducer J. Sound Vib. 505 116144
[8] Liu X C, Feng S, Shang Z X, Jiang Z Q and
Chen X S 2021 Cyclic behaviour of
re-centering energy-dissipation brace with
pendulum Eng. Struct. 242 112511
[9] Ciȩ żkowski M 2021 Dynamic stabilization and
feedback control of the pendulum in
any desired position J. Sound Vib.
491 115761
[10] Sahu S, Pai S, Manjunath N and Balakrishnan J
2021 The lengthening pendulum: adiabatic
invariance and bursting solutions Phys. Open
7 100067
[11] Marr D and Poggio T 1979 A computational
theory of human stereo vision Proc R Soc
Lond B Biol Sci. 204 301–28
[12] Yang L, Wang B, Zhang R, Zhou H and Wang R
2017 Analysis on location accuracy for the
binocular stereo vision system IEEE
Photonics J. 10 1–16
[13] Rodríguez-Quiñonez J C, Sergiyenko O,
Flores-Fuentes W, Rivas-Lopez M,
Hernandez-Balbuena D, Rascón R and
Mercorelli P 2017 Improve a 3D distance
measurement accuracy in stereo vision
March 2022
systems using optimization methods’
approach Opto-Electron. Rev.
25 24–32
[14] Zhang X, Song Y, Yang Y and Pan H 2017
Stereo vision-based autonomous robot
calibration Robot. Auton. Syst. 93 43–51
[15] Ramírez-Hernández L R,
Rodríguez-Quiñonez J C,
Castro-Toscano M J,
Hernández-Balbuena D, Flores-Fuentes W,
Rascón-Carmona R, Lindner R and
Sergiyenko O 2020 Improve
three-dimensional point localization
accuracy in stereo vision systems using a
novel camera calibration method Int. J. Adv.
Robot. Syst. 17 1729881419896717
[16] Ristanto S, Nugroho W, Sulistya E and
Suparta G B 2021 System and method for
stereoscopic image acquisition AIP Conf.
Proc. 2374 020014
[17] Staniek M 2017 Stereo vision method
application to road inspection Balt. J. Road
Bridge Eng. 12 38–47
[18] Song W, Yang Y, Fu M, Li Y and Wang M 2018
Lane detection and classification for
forward-collision warning system
based on stereo vision IEEE Sens. J.
18 5151–63
[19] Fan R 2018 Real-time computer stereo vision for
automotive applications Doctoral
Dissertation University of Bristol
[20] Chang Q and Maruyama T 2018 Real-time
stereo vision system: a multi-block matching
on GPU IEEE Access 6 42030–46
[21] McGuire K, De Croon G, De Wagter C, Tuyls K
and Kappen H 2017 Efficient optical flow
and stereo vision for velocity estimation and
obstacle avoidance on an autonomous pocket
drone IEEE Robot. Autom. Lett.
2 1070–6
[22] Weidner N, Rahman S, Li A Q and Rekleitis I
2017 Underwater cave mapping using stereo
vision IEEE International Conference on
Robotics and Automation (ICRA) Singapore
29 May–3 June 2017 pp 5709–15
[23] Mironov K, Vladimirova I and Pongratz M 2015
Processing and forecasting the trajectory of a
thrown object measured by the stereo
vision system IFAC-PapersOnLine
48 28–35
[24] Gayanov R, Mironov K, Mukhametshin R,
Vokhmintsev A and Kurennov D 2018
Transportation of small objects by robotic
throwing and catching: applying genetic
programming for trajectory estimation
IFAC-PapersOnLine 51 533–7
9 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 2 4
S Ristanto et al
[25] Gintautas V and Dan Hübler A 2009 A simple,
low-cost, data-logging pendulum built from
a computer mouse Phys. Educ. 44 488
[26] Yulkifli Y, Afandi Z and Yohandri Y 2017
Development of gravity acceleration
measurement using simple harmonic motion
pendulum method based on digital
technology and photogate sensor IOP Conf.
Ser.: Mater. Sci. Eng. 335 012064
[27] Alho J, Silva H, Teodoro V and Bonfait G 2019
A simple pendulum studied with a low-cost
wireless acquisition board Phys. Educ.
54 015015
[28] Yang H, Xiao J, Yang T and Qiu C 2011 A
simple method to measure the trajectory of a
spherical pendulum Eur. J. Phys.
32 867
[29] Ladino L A and Rondón H S 2017 Determining
the damping coefficient of a simple
pendulum oscillating in air Phys. Educ.
52 033007
[30] Fowles G and Cassiday G 2005 Analytical
Mechanics 7th edn (Belmont, CA: Thomson
Brooks/Cole)
Sigit Ristanto, MSc, works as
a lecturer at the Department of
Physics Education, Universitas
PGRI Semarang, Indonesia. His
main interest is in the Physics of
Imaging and its use for developing
experimental tools based on real
situations useful in physics education.
He is also a PhD candidature at the
Department of Physics, Universitas
Gadjah Mada, Indonesia.
March 2022
Waskito Nugroho, PhD (RIP).
Previously worked as a lecturer at the
Department of Physics, Universitas
Gadjah Mada, Indonesia. Mainly
concerned with programming and
measurement based on the Physics of
Imaging. His works related to Non-
Destructive Testing, Imaging Physics,
Applied Nuclear Physics, and Laser-
Induced Breakdown Spectroscopy
(LIBS).
Eko Sulistya, PhD, a lecturer at the
Department of Physics, Universitas
Gadjah Mada, Indonesia and also
the Head of the Basic Physics
Laboratory, in charge of carrying
out basic physics experimental
services and developing experimental
equipment and methods. He works
mostly on Computational Physics
and developing educational kits using
sensors and Arduino.
Professor Gede Bayu Suparta,
PhD, works as a full Professor at the
Department of Physics, Universitas
Gadjah Mada, Indonesia. He brings
the idea of the Physics of Imaging
as the implementation of Physics
concepts and methods for visual
inspection, industrial imaging,
laboratory imaging, medical imaging,
digital radiography and computed
tomography.
10 P hy s . E d u c . 5 7 ( 2 0 2 2 ) 0 2 5 0 2 4