Study of a Detection and Recognition Algorithm for High-Voltage Switch Cabinet Based on Deep Learning with an Improved Faster-RCNN
Chenzhao Fu, Wenrong Si, Qiyu Lu, Chun-Bo Shi, Qin-Jian Gao, Hai Wang, Chao Wang · 2018
Automatic detection and recognition of the switches of high-voltage (HV) cabinet is an important problem for on-line inspection of power equipments. Aiming at this problem, this paper introduced a deep learning method of artificial intelligence (AI) and proposed a method based on a faster region-based convolutional neural network (Faster-RCNN) network for detecting the switch targets of HV cabinet. This method can automatically detect a variety kinds of switches from the acquired HV cabinet images. It replaces the fully connection layers of initial Faster-RCNN by convolution layers, at the same time, the edge boxes and the scores are predicted at the most possible location of each target. The proposed method can not only improve the generalization of the network, but also decrease the computational complexity. Thus, accurate and efficient automatic detection of switch targets can be achieved. The application results show that it can better meet the needs of practical engineering applications and provide technical support for automatic monitoring of power equipment.