Fast Detection of Hidden Dangers in Transmission Line Corridors Based on Region Partitioning and Feature Extraction

Abstract

Hidden dangers like large-scale construction machinery are the main causes of line trips. In this paper, we propose a fast detection algorithm for hidden dangers in transmission line corridors based on region partitioning and feature extraction. Since the scenes in the sky are simple and stable compared with the scenes on the ground, we detect hidden dangers in the sky and those on the ground separately. For hidden dangers in the sky, we firstly designed an algorithm to calculate the mask image for sky area. After dividing the sky area, we extracted features including colors and shapes of the difference areas in order to eliminate the interference factors. The targets left are the hidden dangers we truly cared about. For hidden dangers on the ground, firstly we design a pre-processing algorithm to eliminate the influence of uneven illumination on subsequent matching. Secondly, we propose a multi-scale gray-weighted average method to fuse multiple channels of multiple color spaces, by which we can effectively suppress noise caused by camera shake and maximize the area that cannot be matched. Thirdly, we use the Haar feature density map to filter the photo to remove discrete pixel points caused by small disturbances. Finally, we extract multiple features and fuse those features on decision level to obtain the final matching result. When deployed on intelligent monitoring devices and tested with massive scene photos, our algorithm has achieved the expected running efficiency and detection accuracy.