Detection of red light running behavior of two wheeled motor vehicles at intersections

With the rapid development of urbanization in China and the exponential growth of urban population, two wheeled motor vehicles have become a common mode of transportation and are used for short distance delivery. However, due to the lack of safety awareness, the behavior of drivers of two wheel motor vehicles running red lights has become a common problem, posing significant safety hazards to themselves and others. To solve this problem, we first took a typical intersection as the research object, collected 5200 images in this scene, and divided them into 6 categories according to weather conditions. Then, a multi class object detection algorithm (MCODA) based on convolutional neural network (CNN) was proposed, the MCODA can detect signal indicators and two wheeled vehicles in images. Finally, the algorithm for judging a two wheeled vehicle running the red light (TWV-RRL) is proposed. Tested in different scenarios, it was found that the accuracy rate was above 90%, with a comprehensive accuracy rate of 97.3%, achieving relatively accurate detection of red light running behavior by two wheeled vehicles.