Jiajie Jin, Hao Qin, Wei Sun, Yucheng Qian, Haigang Wang, Zhi Wang, Jiyong Zhou, and Xiongfeng Huang
Substation, target recognition, target tracking, monocular ranging
Outdoor substations have repeatedly suffered power failures due to the intrusion of abnormal targets. This not only seriously affects the stability of the power system but also causes huge economic losses. Given the identification and early warning of foreign object intrusion in substations, existing target detection algorithms have shortcomings in recognition time and accuracy, making it difficult to meet the growing requirements of intelligence. As such, the You Only Look Once version 5 (YOLOv5) algorithm is first optimised to improve the accuracy of target recognition. Then the ability to determine the occluded target is enhanced by improving the ByteTrack algorithm, and the problem caused by occlusion in target tracking is effectively solved. Finally, combining the high-altitude and ground ranging methods, the monocular vision technology is used to realise the distance measurement, so as to realise the identification and early warning of abnormal targets in substations. The experimental results show that compared with other algorithms, the proposed algorithm can improve the intelligence level of unmanned substations and provide timely early warning for abnormal targets.
Important Links:
Go Back
