Detection and Segmentation of Power Line Fires in Videos

Poor vegetation management around power lines can cause severe fires that lead to tremendous economic losses, environmental degradation, and fatalities. The early discovery of a fire's presence is the key to avoiding catastrophic damages. In this paper, we propose a hybrid fire detection framework based on a deep convolutional neural network (CNN) and a pixel-based fire detector to automatically detect both the presence of fire and its scale and position information. The pre-trained deep CNN serve as a binary classifier to detect the presence of fire. The pixel-based fire detector is designed to find the fire pixels in the video frames, which indicate the scale and location of the fire. Case studies are carried out on six real-world videos to validate the proposed framework. It is shown that the proposed approach can effectively detect fire and locate the fire pixels in the testing fire videos.