Maintenance Personnel Detection and Analysis Using Mask-RCNN Optimization on Power Grid Monitoring Video

摘要

In recent years, deep learning theory and applications have been grown rapidly. Its application aspects has been widely extended to medical care, unmanned driving, intelligent monitoring and other fields. In this paper, we focus on detecting and analyzing the movements of maintenance personnel based on power grid surveillance videos by using MASK-RCNN. Firstly, we detect the maintenance personnel in the video data using optimized MASK-RCNN network. Then, we plot the corresponding personnel path image using segmentation and centroid detection, which can accurately count the personnel trajectory with in-and-out information. Secondly, this paper introduce a tracking-learning-detection algorithm to further track and analyze interested feature events of power grid video. The experimental results show that our algorithm can accurately detect multiple personnel and obtain the key features of the video contents.