VIDEO PROFILING FOR ANOMALY DETECTION

Abstract

There has been an increasing demand for automatic methods which can analyse huge quantities of video data for surveillance which is continuously generated by closed circuit television (CCTV) Systems. Our main objective of deploying an automated visual surveillance system is to detect abnormal behavior patterns and recognize the normal ones.There are many other methods which solve the problem of anomaly detection in video surveillance.In this thesis, we present a novel method to detect anomalies in a video surveillance system. We will make use of FEM toolbox in order to detect abnormalities in a video sequence. This toolbox will give us modal frequencies of the frames of a video, which will help us to find out as to which actions are periodic and which actions are non-periodic. Also, we will compare periodicities of various periodic actions and compare and contrast two similar actions on grounds of these frequencies.