Abstract:Hazard detection is an important computer vision area in electric power communication network operation and maintenance. In this paper, we propose a hazard detection method in electric power communication network.This method turns the short-range wireless localization information into a confidence mask corresponding to the image size. The mask will be input into different object detection frameworks, including the one-stage detector and the two-stage detector, to improve the accuracy of the detections and alleviate the problems of false positives and false negatives. For both detectors, we use the confidence score revision based on the wireless localization to improve the detection performance. In addition, datasets containing specific scenarios are collected to train and test the proposed method of hazard detection task in electric power communication network. Finally, this paper adopts various metrics to evaluate the effectiveness of the proposed method. The experiments demonstrate the feasibility and superiority of the proposed method in both theory and practice.