Research on Insulator Detection Method Based on Machine Vision
Yu Chen, Jinrong Miao, Chenbiao Yang, Yuting Yan, Zhefei Wang, Runping He · 2023
The detection of insulators is of critical significance because they are the main piece of hardware in overhead transmission lines and are crucial to the transmission and distribution process. However, in the actual test process, the complex environment and limited equipment measurement capabilities make it difficult for the previous target detection methods to meet the demands of insulators with rich and varied scales. For the purpose of above challenges in the field of insulator detection, a new insulator detection algorithm based on machine vision technology was proposed by using computer vision and image processing technology. Firstly, for enrich the insulator dataset, the mosaic data augmentation algorithm was used to improve the background complexity of the image to be detected. Secondly, in order to improve the problem of small dataset, a transfer learning strategy was added to the YOLOV4 network. Finally, the SE attention module was used to improve the performance of insulator identification. Experimental results show that the improved YOLOV4 network model has better results through data augmentation, transfer learning, and attention mechanism in the network, mAP reached 97%, compared with YOLOV4, mAP increased by 19%, and the network can better realize insulator detection.