Remote Surveillance Object Detection Based on YOLO with Gimbal Camera Tracking Control

This research integrates edge computing with the developed YOLO object detection and controls an optical zooming gimbal camera on an unmanned ground vehicle (UGV) to achieve precise target tracking and distance estimation. The system employs YOLO's object detection (OD) algorithm, PID control of the gimbal camera, and optical zooming strategy to enhance the tracking accuracy of an moving object. By utilizing the polynomial curve fitting, it can map and establish the relationship between the target pixels and predicted distances. The automatic zooming technology improves the real distance recognition accuracy and reduces the OD errors significantly. Developed fast frame-per-second YOLO OD method ensures target recognition with varying distances. The OD system on the UGV is suitable for large-scale remote surveillance.