Research on Foreground Extraction and Target Detection and Tracking Algorithm of Surveillance Video Based on Deep Learning
Peilei Ye · 2023
With the popularization of intelligent monitoring system, it is more and more important to detect remnants and crowd anomalies. At present, most of the anomaly detection methods of left-behind objects rely on tracking the object carriers, while the anomaly detection methods of crowds mostly judge the anomaly by detecting and tracking the trajectory of pedestrian targets and then analyzing whether their behaviors conform to the normal state. How to efficiently analyze the video data acquired by the video surveillance network by artificial intelligence is a frontier topic in the field of computer vision recently. Moreover, with the rapid development of computer science and technology and video surveillance hardware, the industry has higher and higher requirements for intelligent surveillance technology. Therefore, the quality standard of video target tracking and detection technology based on deep learning directly affects the quality of video target tracking based on deep learning. This paper mainly introduces three common methods of video target tracking and detection based on deep learning, and makes a comparative study of different methods, analyzing their advantages and disadvantages, hoping to provide some help for the further research of video tracking and detection algorithm.