Professional Documents
Culture Documents
4-1 Introduction
The solution of most photogrammetric problems generally requires some type of photographic
measurement. For certain problems the measurements may simply be the lengths of lines between
imaged points. However, rectangular coordinates of imaged points are the most common type of
photographic measurement, and they are used directly in many photogrammetric equations.
Photographic measurements are usually made on positives printed on paper, film, or glass, or in
digital images manipulated on a computer. They could also be made directly on the negatives;
however, this is seldom done because it can deface the imagery, and it is important to preserve the
negatives for making additional prints. It is common, however, to make digital scans directly from the
negatives, thus avoiding additional expense associated with making positive prints.
Equipment used for making photographic measurements varies from inexpensive, simple scales
to very precise and complex machines that provide computer-compatible digital output. These various
types of instruments and the manner in which they are used are described in this chapter. Because of
several effects, there will be systematic errors associated with practically all photographic
measurements. The sources of these errors and the manners by which they are eliminated are also
discussed in this chapter.
The position of any image on a photograph, such as point a of Fig. 4-1, is given by its rectangular
coordinates xa and ya, where xa is the perpendicular distance from the y axis to a and ya is the
perpendicular distance from the x axis to a. Similarly, the photographic position of image point b is
given by its rectangular coordinates xb and yb.
It is very common for aerial cameras to have eight fiducials installed, in both side and corner
locations. Figures 1-2, 1-6, and 1-7 show this fiducial mark configuration. The photographic
coordinate system in this case is still defined as in Fig. 4-1. Eight fiducials enable somewhat more
accurate corrections to be made for systematic errors in measured image coordinates.
On digital images, coordinates are often expressed as row and column numbers of individual
pixels. If pixel dimensions are known, the dimensionless row and column values can be converted to
linear measurements. Note that for many common image manipulation programs, pixel coordinates
are given in the order of column (abscissa) followed by row (ordinate). This results in a “left-handed”
coordinate system, which may lead to problems in certain calculations.
Rectangular coordinates are a very basic and useful type of photographic measurement. They are
used in many different types of computations. As an example, they can be used to calculate the photo
distances between points by using simple analytic geometry. Photographic distance ab of Fig. 4-1, for
example, may be calculated from rectangular coordinates as follows:
(4-1)
When greater accuracy is desired, a device such as the glass scale of Fig. 4-3 may be used. Glass
scales have fine graduations etched on the bottom surface and are equipped with magnifying eyepieces
which can slide along the scale. Glass scales of the type shown in Fig. 4-3 can be obtained in either 6-
or 12-in length with either millimeter graduations (least graduations of 0.1 mm) or inch graduations
(least graduations of 0.005 in). With a glass scale, readings may be estimated quite readily to one-
tenth of the smallest division, but these scales cannot be used to lay off distances.
FIGURE 4-3 Glass scales for photographic measurements. (Courtesy Teledyne-Gurley Co.)
1. (a) Film distortions due to shrinkage, expansion, and lack of flatness; (b) CCD array distortions
due to electrical signal timing issues or lack of flatness of the chip surface
2. (a) Failure of photo coordinate axes to intersect at the principal point; (b) Failure of principal
point to be aligned with center of CCD array
3. Lens distortions
4. Atmospheric refraction distortions
5. Earth curvature distortion
Corrections may be applied to eliminate the effects of these systematic errors. However, not all
corrections need to be made for all photogrammetric problems; in fact, for work of coarse accuracy
they may all be ignored. If, for example, an engineer’s scale has been used to make the measurements,
uncertainty in the photo coordinates may be so great that the small magnitudes of these systematic
errors become insignificant. On the other hand, if precise measurements for an analytical
photogrammetry problem have been made with a comparator, all the corrections may be significant.
The decision as to which corrections are necessary for a particular photogrammetric problem can be
made after considering required accuracy versus magnitude of error caused by neglecting the
correction.
4-8 Distortions of Photographic Films and Papers
In photogrammetric work, true positions of images in the picture are required. Photo coordinates
measured by any of the previously discussed methods will unavoidably contain small errors due to
shrinkage or expansion of the photographic materials that support the emulsion of the negative and
positive. In addition, since photogrammetric equations derived for applications involving frame
cameras assume a flat image plane, any lack of flatness will likewise cause errors. These errors can
also be categorized as film distortions, and they are generally the most difficult to compensate for,
due to their nonhomogeneous nature. Photo coordinates must be corrected for these errors before they
are used in photogrammetric calculations; otherwise, errors will be present in the computed results.
The magnitude of error in computed values will depend upon the severity of the film distortions,
which depends upon the type of emulsion support materials used and the flatness of the camera platen.
Most photographic films used to produce negatives for photogrammetric work have excellent
dimensional stability, but some small changes in size do occur during processing and storage.
Dimensional change during storage may be held to a minimum by maintaining constant temperature
and humidity in the storage room. The actual amount of distortion present in a film is a function of
several variables, including the type of film and its thickness. Typical values may vary from almost
negligible amounts up to approximately 0.2 percent.
Paper media are generally much less stable than film. Whether the images are photographically
reproduced or are printed from a scanned image, paper copies often require corrections to measured
distances in order to produce satisfactory results.
(4-2)
(4-3)
I n Eqs. (4-2) and (4-3), x′a and y′a are corrected photo coordinates and xa and ya are measured
coordinates. The ratios xc/xm and yc/ym are simply scale factors in the x and y directions, respectively.
This method is also appropriate for images that have been subjected to substantial enlargement or are
being measured with arbitrary coordinates, for example, an image from a digital camera having a CCD
frame width of 20 mm printed at a width of 150 mm on a sheet of paper. If the precise dimensions of
the CCD frame have been determined through calibration, a correction factor based on Eq. (4-2) can
be computed. In another case, a digital image may be imported into a Computer Aided Drafting
(CAD) drawing. In this case the coordinates of the CAD drawing may be completely arbitrary and
need to be related to a calibrated reference in order to obtain correct measurements. Equations (4-2)
and (4-3) can be used by making CAD-unit measurements of the calibrated distances and applying the
correction factors to subsequent measurements.
Example 4-1
For a particular photograph, the measured x and y fiducial distances were 233.8 and 233.5 mm,
respectively. The corresponding x and y calibrated fiducial distances were 232.604 and 232.621 mm,
respectively. Compute the corrected values for the measured photo coordinates which are listed in
columns (b) and (c) in the table below.
Each of the measured values is multiplied by the appropriate constant above, and the corrected
coordinates are entered in columns (d) and (e) of the table.
For high-accuracy applications, shrinkage or expansion corrections may be applied through the x
and y scale factors of a two-dimensional affine coordinate transformation. This method is particularly
well-suited for analytical photogrammetric calculations. This procedure is described in Sec. C-6, and a
numerical example is presented.
(4-4)
In Eq. (4-4), Δr is the amount of radial lens distortion, r is the radial distance from the principal point,
and k1, k2, k3, and k4 are coefficients of the polynomial. The coefficients of the polynomial are solved
by least squares using the distortion values from the calibration report. To correct the x, y position of
an image point, the distance r from the image point to the principal point is computed and used to
compute the value of Δr from Eq. (4-4). This is done by first converting the fiducial coordinates x and
y, to coordinates and , relative to the principal point, by Eqs. (4-5) and (4-6). Then Eq. (4-7) is used
to compute the value of r to use in Eq. (4-4).
(4-5)
(4-6)
(4-7)
After the radial lens distortion value of Δr is computed, its x and y components (corrections δx
and δy) are computed and subtracted from and , respectively. The δx and δy corrections are based
on a similar-triangle relationship, as shown in Fig. 4-4. By similar triangles of that figure
FIGURE 4-4 Relationship between radial lens distortion and corrections to x and y coordinates.
from which
(4-8)
(4-9)
(4-11)
Example 4-2
An older USGS camera calibration report specifies the calibrated focal length f = 153.206 mm and
coordinates of the calibrated principal point as xp = 0.008 mm and yp = –0.001 mm. The report also
lists mean radial lens distortion values given in columns (a) and (b) of the table below. Using these
calibration values, compute the corrected coordinates for an image point having coordinates x =
62.579 mm, y = –80.916 mm relative to the fiducial axes.
Solution Compute r values (in meters) in column (c) by the following equation (see Fig. 4-5):
FIGURE 4-5 Illustration of radial distance r as it relates to focal length f and field angle θ.
Using the least squares method presented in Sec. B-11 (see Example B-6), the following k values were
computed.
Compute the distance from the principal point to the image point, using Eqs. (4-5), (4-6), and (4-7).
Given this value for r and the k coefficients, compute Δr by Eq. (4-4):
Lens design in modern aerial mapping cameras has evolved to such a level that symmetric radial
lens distortion is of the same order of magnitude as decentering distortion, and camera calibration
reports have been adapted to accommodate this change. For example, the mathematical model used in
the current USGS calibration procedure, known as the Simultaneous Multi-camera Analytical
Calibration (SMAC), computes both symmetric radial and decentering distortion parameters directly
by least squares. Principal point coordinates and focal length are also determined in the solution. The
USGS camera calibration report lists polynomial coefficients for symmetric radial lens distortion (k0,
k1, k2, k3, k4), and decentering distortion (p1, p2, p3, p4). It also gives calibrated principal point
coordinates (xp, yp). To compute coordinates (xc, yc) corrected for these systematic errors, the
following equations are used:
(4-12)
(4-13)
(4-14)
(4-15)
(4-16)
(4-17)
In Eqs. (4-12) through (4-17), and are coordinates of the image relative to the principal point
as computed by Eqs. (4-5) and (4-6), respectively; r is the radial distance from the image to the
principal point as computed by Eq. (4-7); k0, k1, k2, k3, and k4 are coefficients of symmetric radial lens
distortion from the calibration report; p1, p2, p3, and p4 are coefficients of decentering distortion from
the calibration report; δx and δy are the symmetric radial lens distortion corrections to and ,
respectively; and Δx and Δy are the decentering distortion corrections to and , respectively.
Example 4-3
The parameters of a current USGS camera calibration report are given in the following table. Using
these calibration values, compute the corrected coordinates for an image point having coordinates x =
–47.018 mm, y = 43.430 mm relative to the fiducial axes.
Compute symmetric radial lens distortion corrections δx and δy, using Eqs. (4-12) and (4-13),
respectively.
Compute decentering distortion corrections Δx and Δy, using Eqs. (4-14) and (4-15), respectively.
Compute the corrected coordinates xc and yc, using Eqs. (4-16) and (4-17), respectively.
I n Fig. 4-6, if a straight path had been followed by the light ray from object point A, then its
image would have been at a′. The angular distortion due to refraction is Δα, and the linear distortion
on the photograph is Δr. Refraction causes all imaged points to be displaced outward from their
correct positions. The magnitude of refraction distortion increases with increasing flying height and
with increasing α angle. Refraction distortion occurs radially from the photographic nadir point1
(principal point of a vertical photo) and is zero at the nadir point. Atmospheric refraction in tilted
photographs is treated in Sec. 10-16.
The usual approach to the atmospheric refraction correction is based on the assumption that
change in the refractive index of air is directly proportional to change in height. Starting with the
incident angle of the light ray at ground level, Snell’s law can be solved continuously along the ray
path for each infinitesimal change in angle due to refraction. When all the infinitesimal changes are
summed, the total is proportional to the tangent of the incident angle. (Derivation of the relationship
involves the solution of a differential equation, which is beyond the scope of this text.) The
proportionality constant is based on the values of the refractive indices at ground level and at the
camera position, which are related to elevation. The relationship that expresses the angular distortion
Δα as a function of α is
(4-18)
In this equation, α is the angle between the vertical and the ray of light, as shown in Fig. 4-6, and K is
a value which depends upon the flying height above mean sea level and the elevation of the object
point. There are several different approaches to calculating a value for K, with most assuming a
standard atmosphere. A convenient method, adapted from the Manual of Photogrammetry, is to
compute K by
(4-19)
In Eq. (4-19), H is the flying height of the camera above mean sea level in kilometers, and h is the
elevation of the object point above mean sea level in kilometers. The units of K are degrees.
The procedure for computing atmospheric refraction corrections to image coordinates on a
vertical photo begins by computing radial distance r from the principal point to the image, using Eq.
(4-20). In this equation, the x and y image coordinates do not necessarily need to be related to the
principal point since the error due to the assumption of vertical photography far overshadows any
error which would be introduced.
(4-20)
(4-21)
The values of K and tan α from Eqs. (4-19) and (4-21), respectively, are then substituted into Eq. (4-
18) to compute refraction angle Δα.
(4-22)
The radial distance r′ from the principal point to the corrected image location can then be
computed by
(4-23)
(4-24)
The x and y components of atmospheric refraction distortion corrections (δx and δy) can then be
computed by Eqs. (4-8) and (4-9), using the values of x and y in place of and , respectively. To
compute corrected coordinates x′ and y′, the corrections δx and δy are subtracted from x and y,
respectively.
Example 4-4
A vertical photograph taken from a flying height of 3500 m above mean sea level contains the image a
of object point A at coordinates (with respect to the fiducial system) xa = 73.287 mm and ya = –
101.307 mm. If the elevation of object point A is 120 m above mean sea level and the camera had a
focal length of 153.099 mm, compute the x′ and y′ coordinates of the point, corrected for atmospheric
refraction.
Subtract the corrections δx and δy from x and y, respectively to obtain corrected coordinates x′ and y′.
In a situation where a point feature is designated to be at the angle point of a painted chevron
target, it is important to use the intersections of the centerlines of the stripes rather than the outside or
inside corner as the point. Edges that may appear clear and sharp at a glance will become blurred and
indistinct when viewed under magnification as shown in Fig. 4-8. In this figure, it would be difficult
to precisely place cross-hairs on the corner of the chevron, but the centerlines of the stripes can be
identified precisely.
FIGURE 4-8 Zoomed-in view of a chevron target with cross-hair at the intersection of centerlines of
stripes.
In many cases it is necessary to measure edges, particularly when mapping features in an area. It
is then important to make a best estimate of where the true edge of the feature is located. A certain
amount of error can be expected in the measurement due to blurred edges. For example, consider
measuring the lengths of the stripes of the chevron in Fig. 4-8. It is very difficult to identify the
precise location of where the paint meets the underlying pavement.
References
Abdel-Aziz, Y. L.: “Asymmetrical Lens Distortion,” Photogrammetric Engineering and Remote
Sensing, vol. 41, no. 3, 1975, p. 337.
American Society for Photogrammetry and Remote Sensing: Manual of Photogrammetry, 5th ed.,
Bethesda, MD, 2004, chaps. 3 and 10.
Brown, D. C.: “Close-Range Camera Calibration,” Photogrammetric Engineering, vol. 37, no. 8, 1971,
p. 855.
———: “Unflatness of Plates as a Source of Systematic Error in Close-Range Photogrammetry,”
Photogrammetria, vol. 40, no. 9, 1986, p. 343.
Forrest, R. B.: “Refraction Compensation,” Photogrammetric Engineering, vol. 40, no. 5, 1974, p. 577.
Fraser, C. S.: “Photogrammetric Measurement to One Part in a Million,” Photogrammetric
Engineering and Remote Sensing, vol. 58, no. 3, 1992, p. 305.
Fritz, L. W.: “A Complete Comparator Calibration Program,” Photogrammetria, vol. 29, no. 4, 1973,
p. 133.
Frost, R. M.: “Improved Well Positions for Geoscientific Applications: Exploiting NAPP Photographs
with Digitizer and PC-Based Bundle Adjustment Program,” Photogrammetric Engineering and
Remote Sensing, vol. 61, no. 7, 1995, p. 927.
Gyer, M. S.: “Methods for Computing Photogrammetric Refraction Corrections for Vertical and
Oblique Photographs,” Photogrammetric Engineering and Remote Sensing, vol. 62, no. 3, 1996, p.
301.
Jaksic, Z.: “Deformations of Estar-Base Aerial Films,” Photogrammetric Engineering, vol. 38, no. 3,
1972, p. 285.
Light, D. L.: “The New Camera Calibration System at the U.S. Geological Survey,” Photogrammetric
Engineering and Remote Sensing, vol. 58, no. 2, 1992, p. 185.
Murase, T., M. Tanaka, T. Tani, Y. Miyashita, N. Ohkawa, S. Ishiguro, Y. Suzuki, H. Kayanne, and H.
Yamano: “A Photogrammetric Correction Procedure for Light Refraction Effects at a Two-
Medium Boundary,” Photogrammetric Engineering and Remote Sensing, vol. 74, no. 9, 2008, p.
1129.
Scarpace, F. L., and P. R. Wolf: “Atmospheric Refraction,” Photogrammetric Engineering, vol. 39, no.
5, 1973, p. 521.
Scholer, H.: “On Photogrammetric Distortion,” Photogrammetric Engineering and Remote Sensing,
vol. 41, no. 6, 1975, p. 761.
Tayman, W. P.: “User Guide for the USGS Aerial Camera Report of Calibration,” Photogrammetric
Engineering and Remote Sensing, vol. 50, no. 5, 1984, p. 577.
Wiley, A. G., and K. W. Wong: “Geometric Calibration of Zoom Lenses for Computer Vision
Metrology,” Photogrammetric Engineering and Remote Sensing, vol. 61, no. 1, 1995, p. 69.
Problems
4-1. Assume that photo coordinates of points a and b of Fig. 4-1 are xa = 57.62 mm, ya = 67.15 mm,
xb = 22.07 mm, and yb = –49.50 mm. Calculate photo distance ab and radial distances oa and ob.
4-2. Repeat Prob. 4-1 except that the photo coordinates are xa = –71.66 mm, ya = 47.82 mm, xb =
81.43 mm, and yb = 6.05 mm.
4-3. Name and briefly describe the various systematic errors that may exist in photographic
coordinates.
4-4. Calculate the acute angle (to the nearest 0.001°) of intersection of fiducial lines for a camera of
the type shown in Fig. 4-1 if comparator measurements of the fiducial marks on a calibration “flash
plate” were as follows:
4-5. Repeat Prob. 4-4, except that the following flash plate measurements were taken:
4-6. If the intersection of fiducial lines of the camera of Prob. 4-4 defines the principal point
exactly, what are the x and y photo coordinates of the four fiducial marks in the photo system?
Assume that the x and y photo coordinate axes are parallel to the comparator axes.
4-7. Repeat Prob. 4-6, except that it applies to the data of Prob. 4-5.
4-8. On a paper-print positive, the measured x distance between fiducials (1 and 3) was 224.9 mm
and y between fiducials (2 and 4) was 223.5 mm. These x and y distances determined in camera
calibration were 225.433 and 225.693 mm, respectively. Using the method of Example 4-1, calculate
shrinkage-corrected coordinates of points a, b, and c whose coordinates were measured on the paper
print as follows:
4-9. Repeat Prob. 4-8, except that the measured x distance on a paper-print positive between
fiducials 1 and 3 was 225.64 mm and y between fiducials 2 and 4 was 225.83 mm; the calibrated
distances between these same fiducials were 224.282 and 224.155 mm, respectively; and measured
photo coordinates of points a, b, and c were as follows:
4-10. A recent USGS camera calibration report yielded the parameters given in the following table.
Using these calibration values, compute the corrected coordinates (to the nearest micrometer) for an
image point having coordinates x = 73.952 mm and y = 98.155 mm relative to the fiducial axes.
4-11. Repeat Prob. 4-10, except that the coordinates, with respect to the fiducial system, of the image
point are x = –78.090 mm and y = –57.688 mm.
4-12. The photo coordinates listed below have been corrected for film and lens distortions. The
camera that took the photography had a focal length of 151.942 mm, the flying height above mean sea
level (MSL) was 8530 m, and the average elevation of the terrain was 310 m above MSL. Calculate
the photo coordinates (to the nearest micrometer) corrected for atmospheric refraction using the
method of Example 4-4.
4-13. Repeat Prob. 4-12, except that the camera lens had a focal length of 88.916 mm, flying height
above MSL was 4250 m, the average terrain elevation was 170 m above MSL, and the photo
coordinates were as follows:
_____________
1 The photographic nadir point is defined in Sec. 10-3.