SMART TRAFFIC SIGNAL CONTROL BY USING ARTIFICIAL INTELLIGENCE

Project Title

SMART TRAFFIC SIGNAL CONTROL BY USING ARTIFICIAL INTELLIGENCE

Project Area of Specialization Artificial IntelligenceProject Summary

In this modern world, vehicular traffic congestion is an increasingly growing issue. The rise in vehicles purchased each year in no way decreases the number of vehicles on our highways. There is therefore a need to devise a system, especially at intersections, to ensure the smooth flow of vehicles. Previous Traffic schemes were not properly capable of tackling this growth of road traffic. The implementation of smart traffic lights using artificial intelligence and IoT to effectively control all traffic situations is discussed in this paper. Traffic optimization is accomplished by the use of IoT platforms to assign various times to each or all traffic signals, consistent with the number of vehicles on the route, for economical use. As input from cameras, the model program takes traffic density that is abstracted from the digital image method technique. To predict traffic density, an algorithm is used to attenuate traffic jams and if any intersection detects an emergency vehicle, then adjust the algorithm and operate according to cases. The key benefit of this method is that, depending on the traffic situation, it will eliminate the additional waiting for traffic. The video-based traffic flow detection system has become increasingly stable, real-time, and intelligent to further assess the efficiency of the rapid growth of computer vision and digital image processing technology.

Project Objectives

The objective of this smart traffic signal project is to reduce traffic congestion and improve the efficiency of the existing automatic traffic signaling system. The system will use image processing to control the traffic with the help of continuous images, real-time data, and objects on the single lane and according to that system, it will calculate each time change automatically depending on traffic load on the road. The system will function based on the traditional system along with automatic signaling. The system will have an artificial computer vision with the help of a camera. The old traditional traffic light system work on timers that are hardcoded programmed or sensor-based and they need high maintenance vehicles. So, the vehicles on the road follow the lights and the flow of traffic totally depends on that traditional programmed system. The aim of this report is to make the traffic light system run according to the flow of traffic instead of the flow of traffic not run according to traffic lights.

Project Implementation Method

Our project is AI-based and for that, we have developed a prototype with toy cars and ambulances. We have used A4-tech cameras to collect 3200 images of vehicles and annotated these images. Furthermore, we employed the YOLO v3 model for the detection and recognition of vehicles. We have used 3000 images for training and 200 images for training. We have also employed an algorithm to detect and recognize emergency vehicles, such as ambulances and police vehicles and our model provides preference to these emergency vehicles. Our model uses variable waiting time at signals depending on the density of traffic. 

Benefits of the Project

The benefit of this project is basically to try to controls and maintains the traffic signals smartly. Our project will mainly focus on the following objectives:

1. By this, we can help our country by solving one of the biggest problems.
2. We try to monitor and control the traffic by continuously capture images with the help of a CCTV video camera which will be fitted on Signals. By taking real-time pictures of traffic we can reduce the flow of traffic jams.
3. It can eliminate the extra waiting of traffic according to the traffic situation.
4. Develop a system that automatically allots time to traffic signals according to the density of traffic.
5. It is easy and economical to install video cameras. Besides, it would never damage the road, nor would it block the traffic.
6. With the fast development of image processing technology, the video-based traffic flow detection system has become increasingly robust, Real-time, and intelligent.
7. It’s very effective for emergency services and if any intersection detects an emergency vehicle, then adjust the algorithm and operate according to cases.

Technical Details of Final Deliverable

We utilized various instruments and innovations for the improvement of this task. All instruments have done diverse work like cameras (a4-tech webcams) and picture processing which helped us to find out mostly congested roads to control the flow of traffic and help people out with an emergency. We have captured 3300 images and annotated them with 15000 boxes from cameras mounted on traffic poles to differentiate between normal cars and emergency vehicles with the help of Python codes and we did classification for the vehicle type that which vehicle is normal and which vehicle is used for an emergency. We have used YOLO version 3 and train our model on the custom data set. For installing OS on Raspberry Pi, we have used an SD card, mouse, and keyboard. And HDMI cable or VGA cable to HDMI converts. Then we have installed Pycharm on raspberry pi for further implementation.

Final Deliverable of the Project HW/SW integrated systemCore Industry TransportationOther IndustriesCore Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Sustainable Cities and CommunitiesRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	43300
RASPBERRY PI	Equipment	1	13000	13000
A-4 Web CAMERA	Equipment	4	2600	10400
SDCARD	Equipment	2	4300	8600
Toy cars	Equipment	20	40	800
Prototype Frame	Equipment	1	2500	2500
RASPBERRY PI CASING	Equipment	1	500	500
Others	Miscellaneous	1	7500	7500