Insulator Detection Method Based on Improved Faster R-CNN with Aerial Images
Weikuan Lu, Zhili Zhou, Xiukai Ruan, Zhengbing Yan, Guihua Cui · 2021
Insulators are a critical component in power transmission, and the detection of insulators is an essential prerequisite for achieving insulator status and fault diagnosis. An insulator detection method in aerial images based on the improved Faster R-CNN is proposed to address the problems of inaccurate localization and undetected error in the detection of insulators. In this method, the generalized intersection over union (GIoU) is adopted to overcome that the detection is sensitive to various scales insulators in aerial images, and it also improves the accuracy of insulator localization effectively. Meanwhile, the soft non-maximum suppression (Soft-NMS) algorithm is adopted to avoid missing detection of insulators in the post-processing stage because of mutual occlusion in aerial images. The experimental results show that the proposed method can effectively detect insulators in aerial images with complex backgrounds, and the average accuracy is significantly improved compared with different methods.