Integrating Object Detection and Optical Flow Analysis for Real-time Road Accident Detection

This paper proposes an innovative approach for enhancing road safety using improved traffic surveillance, which uses deep learning and computer vision techniques to detect car accidents. By combining the Lucas-Kanade approach and the YOLOv4 model, the system demonstrates practical application in real-world traffic monitoring, as well as effectiveness under a variety of testing scenarios. Despite issues with camera angles and quality, the research opens the door to future improvements, such as the use of more advanced algorithms to overcome present limitations and expanding the variety of traffic scenarios that automated surveillance systems may address. This contributes to the smart transportation systems domain by providing a new perspective on traffic safety management through technology innovation.