An Invasive Target Detection and Localization Strategy Using Pan-Tilt-Zoom Cameras for Security Applications
J. Y. Hu, Chuanshen Zhang, Sheng Xu, Chunjie Chen · 2021
The widely applied security monitoring system mainly relies on human beings and lacks intelligence and flexibility. This paper designs a practical security monitoring system to quickly detect and localize the invasive animals, which includes the YOLOVS, DeepSORT, and proportional-integral-derivative (PID) methods. The proposed system has two parts, i.e., (1) the target identification and detection part, and (2) target localization part. The identification and detection part detects the suspicious invasive target. To resolve the inaccurate and time-lag problems in the target recognition, a strategy combined with YOLOV5 and DeepSOPT is developed. Then, multiple pan-tilt-zoom (PTZ) cameras in the target localization part will keep their lens to follow the target. The target position can be calculated through the rotation angle of the PTZ modules. Furthermore, an improved PID controller using particle swarm optimization (PSO) is proposed to control the PTZ for target localization. The simulation results demonstrate that the propose system can quickly detect the invasive target, and accurately localize the target.