Research on The Application of Object Detection and Tracking Technology in Abnormal Event Detection on Highways

Real-time detection of abnormal events on highways, such as pedestrian intrusion, abnormal vehicle parking, and vehicle converse running, is one of the crucial measures for reducing traffic accidents on highways. However, the current method of detecting abnormal events on highways based on manual inspection of surveillance videos is beset with high costs and high rates of missed detections. With the advancement of deep learning and computer vision technology, abnormal events detection method for highways based on object detection and tracking technology has been proposed, but it still faces the challenge of low detection accuracy and poor application effects in real-world scenarios. To address this challenge, this paper proposes a three-stage based pedestrian intrusion detection method for highways, which achieves a precision of 96.8% and a recall rate of 99.69% in highway scenes. In terms of abnormal vehicle behavior detection on highways, a vehicle tracking algorithm A-BoT-SORT is proposed by introducing the feature detection description algorithm AKAZE based on BoT-SORT algorithm. Experimental results demonstrate that the proposed A-BoT-SORT algorithm has higher accuracy compared with BoT-SORT algorithm.