Research and application of substation intelligent inspection technology based on multi spectral image recognition

Abstract

The traditional inspection method of substation equipment mainly relies on human resources, which has shortages as high work intensity, low work efficiency and lack of real-time. In this paper, the framework of substation whole area intelligent inspection system is designed, which combines the real-time performance of fixed camera with the mobility of inspection robot, complements each other's advantages. On this basis, through the combination of the improved deep learning Fast-RCNN algorithm and the effective part of the traditional algorithm, the intelligent recognition of typical appearance defects of substation equipment is realized. The feature point detection technology is applied to the meter reading recognition algorithm to improve the adaptability and accuracy under different shooting angles and different shooting conditions. An infrared image target detection algorithm based on deep learning is proposed, which can realize the intelligent extraction of the temperature information of the focus object and intelligent diagnosis of overheating defects without the influence of the accumulated error of camera head. The tracking and positioning of personnel in the substation is realized by the combination of measurement learning, Kalman filter and camera calibration technology, and then realize the real-time monitoring and intelligent warning of equipment and personnel status in the substation. The system has been piloted in a 200kV substation in Zhejiang Province, which effectively promotes the intelligent level of substation safety management and control.