Video Anomaly Detection in Confined Areas

This paper proposes a new supervised algorithm for detecting abnormal events in confined areas like ATM room, server room etc. In the training phase, algorithm learns the motion path and speed of objects in the video. In the testing phase, if any motion happens other than in the learned motion path or the speed of object has large variation from the learned speed then the algorithm alert it as abnormal event. The proposed method process video in groups of frames. The algorithm uses statistical functions to learn the motion path and speed of objects in a video.