International Journal of Advanced Technology in Engineering and Science www.ijates.com Volume No.02, Issue No. 12, December 2014 ISSN (online): 2348 – 7550 823 | Page USING SELF ORGANIZING MAP FOR ABNORMAL BEHAVIOR DETECTION IN VISUAL SURVEILLANCE Ms. Pratibha S.Rane 1 , Ms. Prarthana M. Modak 2 , Ms. Nilam N. Shaikh 3 1, 2, 3 Computer Engineering Department, S.S.P.M. College of Engineering, Kankavli, Mumbai University, (India) ABSTRAT Visual surveillance is an active research topic in the field of computer vision. The task of visual surveillance system is to automatically track object in video image sequences and monitor their activity. Neurobiological studies have concluded that the human brain can perceive actions by observing only the human body poses (postures) during action execution. Thus, actions can be described as sequences of consecutive human body poses, in terms of human body silhouettes. In this work Self organizing map (SOM) is used to find basic posture prototype of action in the video frame. SOM is a type of artificial neural network that is trained using unsupervised learning to produce a low – dimensional, discretized representation of the input space of training samples, called a map. Behavior classifier then detects abnormal behavior by monitoring continuous frames and helps to detect suspicious activity in visual surveillance method. Keywords- Video Surveillance Systems, Anomaly, SOM, Classifier. I. INTRODUCTION Video surveillance systems are to be seen to monitor busy environment like shopping malls, financial institute etc. The videos from these systems are later analyzed to find out any crime, or in some cases human operator continuously keep watch on these video so that he can detect if something abnormal happens. But because of human limitation and huge amount of data it is possible that person get bored which affect on accuracy of system. We proposed a system that can automatically detect abnormal behavior. Neurobiological studies have concluded that the human brain can perceive actions by observing only the human body poses (postures) during action execution. Thus, actions can be described as sequences of consecutive human body poses, in terms of human body silhouettes [1]. In this work Self organizing map (SOM) is used to find basic posture prototype of action in the video frame. Posture of a person provide clue for the understanding his activity. The Self-Organizing Map (SOM) developed by prof. Kohonen is used here. The SOM belong to category of competitive network; where no human intervention is needed during learning [6]. One of the important characteristic of SOM is „topology preserving‟ property i.e. points that are near each other in the in put space are mapped to nearby map units in the SOM. This research focus on detection of anomalous behavior in an environment through the use of a prototype based video surveillance system. The system developed can only handle a subset of possible events in an environment and by no means is it a fully fledged surveillance system, but instead it is a tool that is used as concept demonstrator. The system designed considering only one person in current context with static background. The position of camera, picture quality, shadow may limit accuracy of the anomaly detection system.