Research on Lightweight Power Surveillance Video Key Frame Recognition Method for Edge Computing

Power system video surveillance is an important means to ensure the safe operation of power equipment, but the massive video data generated by it brings great pressure to edge devices. Traditional video processing methods often use cloud computing mode to transmit data to remote servers for processing, which not only increases the burden of network bandwidth, but also brings a large transmission delay. In the edge computing mode, although localized processing can be realized, limited by the computing power of edge devices, how to achieve efficient video processing under resource-constrained conditions has become an urgent problem to be solved. In this paper, we propose a lightweight keyframe recognition method for power surveillance video suitable for edge computing environment. By using sliding window mechanism and depthwise separable convolutional network, combined with adaptive resource adjustment strategy, this method achieves efficient key frame recognition under the condition of limited computing resources. Experimental results show that compared with traditional methods, the proposed method significantly reduces computing resource consumption and average processing delay while maintaining high recognition accuracy, and improves the processing efficiency in the edge computing environment.