VIDEO ANALYSIS AND ANOMALY DETECTION USING FUZZY LOGIC

Abstract

Recognition and understand of human activity in video have gained considerable research attention due to the potential application in various domains [4], such as video surveillance and monitoring, human-computer interfaces, content-based video analysis and behavioural biometrics. In video surveillance, the main objective is to be able to detect events of interest to aid security personnel. Video surveillance is done to find anomalies i.e. unwanted behaviour, in the video data. In this dissertation we propose an approach to find the anomalies in the video data using fuzzy logic. Our work is divided in to two parts, first part tracks multiple objects in the video using kalman filter. All the moving objects in the video are detected in this step. Second part finds the trajectory of each detected object and matches these trajectories with those in the training matrix using fuzzy logic. So we find if the trajectory of the object in normal or anomalous. If the trajectory is anomalous, then an alarm is set to inform about the occurrence of anomalous event. There is no need for any manual control by human beings. All the work is done automatically. Human operator is informed about the anomaly by the alarm which is set when some anomalous event occurs.