High-efficiency insulator detection system in complex space environment

• Main Authors: ZHONG XUKE, ZHONG YUZHONG, TSUKUDA MATSUYOSHI
• Format: Patent
• Language: Chinese; English
• Published: 26.04.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention belongs to the field of insulator detection in machine vision, and provides an insulator high-efficiency detection system in a complex space environment, which comprises a backbone network used for carrying out feature extraction on an input image containing an insulator to obtain a feature map; the region recommendation module is used for recommending interested regions on the feature map; the bilinear pooling module is used for pooling the recommended interested regions to obtain region feature maps with the same size; and the Transform model is used for carrying out local region feature weighting on the region feature maps with the same size, and outputting results to carry out insulator position prediction and classification prediction by using corresponding full connection layers respectively. According to the method, on the basis of keeping the advantage of high recall rate of the Faster R-CNN algorithm, the advantages of the convolutional neural network and the Transform model are combine