Traffic Rules Violation Detection System
Dnyanendra Jagadish Borase, Girish Vilas Shimpi, Mahesh Gopal Chauthe, Mayur Dilip Patil, Anup Dange
Computer Engineering GHRCEM
Pune, India
Abstract—In the world of new developing technologies, traffic violations have become a critical problem for most developing countries. As the population grows, the number of vehicles on the roads also increases rapidly and traffic violations also increase exponentially. The world is rapidly urbanizing. It has led to a multi fold increase in the number of vehicles driving on city roads that causes traffic violations more critical these days. This causes serious destruction property and more accidents that can threaten lives people. Solve the alarming problem and prevent it unpredictable consequences, detection of traffic violations systems are needed. Therefore, managing traffic violations has become a tedious task. Although there are several automated technologies to manage traffic violations, due to uneven lighting conditions, the variety of license plate formats makes it very challenging to manage these conditions. So, the solution to this problem is to develop a system that is linked to several parameters such as traffic signal detection, speed estimation to find out how often the driver violates the traffic rules. This system detects violations of traffic signs, which can be combined with speed estimation to control speeding and this information is sent to a database where the relevant authorities can take necessary action against violators.
In this paper, we present a method that can automatically detect bike-riders without helmets using surveillance videos. The system first uses object segmentation and background subtraction to identify bike riders, and it then determines if they are using a helmet or not. In addition, we introduce a consolidation approach that can improve the accuracy of the proposed system for violation reporting. We tested the three feature representations of this method by comparing their performance. The results of the evaluation revealed that the detection accuracy of the system was
93.80 percent. The proposed method is significantly less expensive and can perform in real-time with an average processing time of 11.58 milliseconds.
KeyWords:-Data Collection, Python Open CV, Object Detec- tion,Tensorflow.
