You are on page 1of 11

Artificial Intelligence In

Smart Traffic Management

Subject : Artificial Intelligence in


Data Science

Presented By:
Sangam Wavre
1st Year ME | Data Science
Pune Institute Of Computer Technology
Index

Introduction

Methodology

Working

Conclusion

References

2
Introduction

Manual traffic systems are insufficient due to
rapid urbanization.

Central Monitoring Systems scalability issue as
increasing cameras.

Solution can be construct new lane which is
expensive and not feasible.

Congestion causing loss of time, fuel which
directly to money and increasing in pollution.

Use of AI in Smart traffic to regulate and govern.

AI function is to perform basis of observation and
experience.

3
Methodology

Cameras on the crossroads which will detect
it with the help of image processing and
getting the count of vehicle will help to
detect the heavy lane.

After detecting the heavy lane, the timer for
that particular will increase and the other
lanes timer will be balanced accordingly.

Python used as programming language.

Algorithms such as blob detection and color
threshold are implemented.

SkLearn library used for machine learning
algorithms.

4
Hardware Used:

Camera used as sensing device.

Traffic lights used as actuators.

Computers will perform processing of Data received from camera and command to Traffic
lights

Software Used :

Pandas : Used to read data from file.

Numpy : Transform multidimensional data as per requirement.

Matplotlib : Used to observe the data how it is scattered.

OpenCV : Used for video analysis, camera calibration, 3d reconstruction and
object detection.

time : Operations related to time.

SkLearn (Sci-Kit Learn) : Used for machine learning algorithms.

imutils : To make basic image processing functions such as translation, rotation,
resizing, skeletonization

Congestion estimation is done by blob detection analysis, which will count the
number of vehicles and compare it with traffic verge.

5
Working

Video taken from camera, OpenCV is used to read that video.

Video converted into frames using CV_CAP_PROP_FRAME_COUNT function of OpenCV.


Each Frame converted into RBG using COLOR_BGR2RGB function to avoid errors.

Network is loaded containing pre-trained weights and configuration file.

Blob is constructed using cv2.dnn.blobFromImage for object detection, associated probabilities and
allotting bounding boxes.


Greater probability is selected from detected bounding boxes.

dlib, the object correlation tracker is used identify the location of the object in upcoming frames.

So the count of the vehicle will be received.

cv2.imshow() is a function used to display an image in a window. 6

Until now vehicles from one video is
detected same steps will be carried out for
each side of cameras.

Video of each side will be passed one by
one in a loop, and the count of vehicle
detected will be stored in a list based on
that count timer in the next turn will be
adjusted.

Lane with more vehicles will get more time
compared to lane with fewer vehicles.

The minimum time limit will be 10 s and ➢
After getting the counts, they are inserted in
the maximum time given to lane will be 40 database so it can be used in future of data
s every time the video will be captured, and analytics and timer prediction. The larger the
based on previous data the timer will be dataset, accurate the result will be.
adjusted.

7
Blob Detection

Binary Large Object and refers to the connected pixel in the binary image.

The term "Large" focuses on the object of a specific size, and that other
"small" binary objects are usually noise.

One of the first and also most common blob detectors is based on the
Laplacian of the Gaussian.

BLOB extraction :
Blob extraction means to separate the BLOBs (objects) in a binary image.
Algorithms are the Recursive Grass-Fire Algorithm, the Sequential Grass-
Fire Algorithm

BLOB representation :
BLOB representation is simply means that convert the BLOB into a few
representative numbers.

BLOB classification :
Is to determine whether constructed blob vehicle or not based on earlier
prototype.

8
Conclusion

Traffic monitoring is the much necessary action for reducing human
interference.

Software advancements perhaps play the most vital role in advanced traffic
management with their ability to analyze the various data input and provide
insights on traffic decreasing and prevention recommendations.

This system analyzes the current of traffic and makes it more convenient for
commuters to reach faster when more traffic.

9
Reference


Aayush Sukhadia, Khush Upadhyay, Meghashree Gundeti, Smit Shah and Manan Shah,
“Optimization of Smart Traffic Governance System Using Artificial Intelligence”,
Journal of Augmented Human Research, Vol. 5, Mar. 2020.
https://link.springer.com/article/10.1007%2Fs41133-020-00035-x

https://en.wikipedia.org/wiki/Blob_detection

10
11

You might also like