Key frames extraction for train roof video based on saliency detection

Abstract

Roof monitoring of running trains can effectively monitor the key components of the roof. In order to check the status of key components on the roof, it is necessary to check the roof surveillance video, which is often time-consuming and labor-intensive. In order to improve the efficiency of status inspection of key components, this paper proposes a key frame extraction algorithm for roof video surveillance based on saliency detection. The algorithm first designs a saliency detection network, which is used to extract candidate video frames containing key components. Then, through feature clustering analysis, the candidate video frames are classified into frame sequences containing components attached different classes, and finally the correlation is calculated for all frames in each frame sequence and the frame with the largest correlation corresponds to the key frame of the current frame sequence.