Computer Vision-based Property Security System

The theft is a multiple crime behavior progressing with human society, and is the primary cause against the current social order and property safety of citizens. Crime data over the years show that, theft occupies the highest proportion in various types of crimes, and burglaries is accounted for more than sixty percent in all the theft case, causing enormous damage to property every year. How to monitor prevent, rapidly process such criminal behavior, minimize the loss of the social property is an urgent task of security department.With the rapid development of video monitoring technology, museums, shopping malls, supermarkets and other indoor open places all adopt video surveillance systems to prevent from the theft behavior. However, the existing video surveillance system can not show all the scenes in display screen without redundant, always has the visual dead angle; In addition, the monitoring system lacks of intelligent analysis and process of the captured video data, then the system need to equip with security men monitoring screen all-weather, but it also bring the problem of low efficiency, high cost, undetected and error detection caused by human errors. To overcome the above shortcomings, this paper designs a property protection system based on computer vision which applies omni-directional vision sensors to monitor property in an open place with no dead angle, simultaneously achieves exceptional theft event detection and confirmed the theft identity relying on PTZ (Pan, Tilt and Zoom) algorithm to. Exceptional theft in video surveillance is the protected property being moved out of the scene without permission, this event will result in color, edge information change in the region where the property is in. It can be seen, the elements composing the theft event are, the thief who implements the behavior, and theft objects which are stolen, and the theft behavior what is abnormal the thief issued to the property. Theft exception is the prerequisite to judge the theft events, and theft object is an essential element to trigger theft exception. So this paper proposes a theft exception detection system based on temporary static region detection algorithm merged with classification process which can distinguish object or human, the theft object or the carryover object.Secondly, after the theft exception occurs, the theft event detection will shift focus on the confirmation of the theft identity. In order to quickly recover the stolen objects need detect and track the thieves to get the detail features. This paper adapts the reconnaissance logical reasoning pattern to confirm theft suspects. Obviously only the person who appears at the same time and the same place with the theft event may implement the theft behavior. So this paper initially confirms theft suspects according to the time and place information of the theft event, then store them in the suspects list."Carrying object detection module"successively judge the suspects in list whether carries theft objects or not. If the carried object is detected as theft object, the person can be confirm as the sole thief, otherwise multiple suspects are confirmed.Lastly, design a device for capturing thief detailed features. In the opening scene in order to quickly find and get the details of the thief features, human generally carry out the following procedure: First, scan all objects in the global scope from macroscopic view.Then target the location of the object, and turn the eye to the location, from the microscopic-view to obtain information of characteristics of human object. According to this ideal, this paper designs a multi-sensor fusion visual device to capture the characteristics of the theft for the police follow-up evidence.This paper develops a computer vision-based property security system, details the implementation of each module, and conducts the relevant experiments in a simulated environment. Experimental results show that the method used in this paper can effectively protect property and provide valuable information for recovery after theft event.