An Automatic Object Detection and Tracking Method Based on Video Surveillance
Enzeng Dong, Yue Zhang, Shengzhi Du · 2020
For intelligent video surveillance, this paper proposes an automatic object detection and tracking method. In proposed method, a cooperative working mechanism between the detector and the tracker is designed, which can initialize the position of the object automatically in the first frame and improves the detection accuracy. This method consists of three modules: detection, tracking and decision-making. The detection module is designed to quickly extract specified categories of objects. The tracking module is used to perform data association and processing based on the kernel correlation filters tracker (KCF). For the object occlusion problem, the decision module can make a judgment and outputs the result. The experimental results show that the proposed method has high real-time performance and robustness, and is suitable in long time video surveillance.