Vision-based highway traffic accident detection
Peng Wang, Cui Ni, Kefeng Li · 2019
The highway traffic scene is relatively simple, the traffic flow is small but the speed of traffic is fast. The traffic accidents on highways are sudden and harmful. Based on highway traffic monitoring video, this paper uses machine learning to accurately detect vehicles on the highways and extract the vehicle trajectory. Then, the background extraction algorithm is used to extract the highway lane boundary lines. According to the change of vehicle trajectory and the position relationship between vehicle trajectory and lane boundary lines, the occurrence of highway traffic accidents can be detected. The experimental results show that the proposed algorithm can achieve good results in traffic accident detection rate and false alarm rate, and can accurately detect the occurrence of highway traffic accidents under different lighting conditions and traffic environment with strong robustness.