Traffic Rule Violation Detection System using Surveillance Videos in real-time

Abstract

In many countries, especially in India, two-wheeler vehicles are the most common mode of transportation. However, because of the lack of security, the danger is extremely high. Governments are working on imposing strict rules and law-abiding behavior in traffic to minimize the risk and injuries. There is a lack of detailed data on the safety-critical behavioural metric of motorbike helmet usage, especially in developing countries like India, where motorcycles are the primary mode of transportation. Due to a lack of evidence, tailored compliance and outreach campaigns, which are critical for accident prevention, are not possible. Despite this, there is a growing need for a secure and timely intelligent device that does not rely on a human observer to detect helmet use by motorcycle riders. As a result, we used a deep learning technique to build a model that can automatically detect the rider's use of a motorcycle and a helmet from video data. Since wearing a motorcycle helmet reduces the risk of fatal injury to motorcycle riders in road traffic accidents by 42 percent, the government has enacted legislation making helmet use mandatory for bike riders. The key goal of our work is to reduce the likelihood of motorcycle riders being killed in traffic accidents. The algorithm's solution is Convolutional Neural Networks (CNN). It's a kind of deep neural network that's often used for visual imagery. The suggested method first determines whether the car is a motorcycle or not. It then decides whether the cycle rider is wearing a helmet or not and whether or not using optical features. On real-world surveillance evidence, the experimental results show that helmet detection accuracy is about 85%