Image Processing based Traffic Surveillance

Motorcycles being primary modes of road transportation globally, face significant risks when traffic rules such as failure to wear helmets, engaging in triple riding, or using mobile phones while driving are disregarded, increasing the likelihood of accidents and injuries, posing threats not only to those directly involved but to the entire community sharing the road. Traditional traffic law enforcement encounters challenges like resource constraints, inconsistent enforcement, limited coverage, and safety concerns for law enforcement officers. To tackle these issues, efforts are underway to revolutionize traffic law enforcement through an automated surveillance system driven by image processing technology. This proposed system extracts the license plate information of the bike when traffic rules violation such as triple riding, failure to wear helmet and mobile phone usage while riding occurs. This proposed system utilizes traffic videos as inputs and employs YOLOv8, an object detection model, to accurately identify bikes and subsequently trained to detect violations. Upon violation detection, an OpenALPR model extracts license plate of the bike and stores this secure data in the Firebase Cloud. This system is tested on Tumkur city road traffic and in college campus for $\mathbf{5 0}$ videos of total duration around 100 hours. The system achieves an accuracy of $\mathbf{9 9 \%}$ in detecting bike riders, helmeted riders with 99 % accuracy, no helmeted riders with 98 % accuracy and detecting the instance of mobile phone usage with an accuracy of 98 % and triple riding instance with 96 %. The system achieved an overall accuracy in detecting the traffic offenders with $\mathbf{9 7 \%}$. Also the system detects the license plate information of the violated vehicle with an accuracy of $\mathbf{9 3 \%}$. Such technology aids law enforcement in consistently enforcing traffic rules, thereby promoting enhanced road safety and orderliness, with the ultimate goal of creating safer roads for everyone.