Research On Abnormal Behavior Recognition In Intelligent Surveillance

Intelligent video surveillance is a cutting-edge in area of computer vision, and it has an extremely important scientific significance and broad application prospects. Compared with traditional video surveillance systems, intelligent video surveillance system acquires the advantages of long-time monitoring, high precise alarm and fast response without human intervention. With the ever increasing challenge of global counter terrorism and the more complex public safety, intelligent video surveillance has become a powerful auxiliary tool for terrorist attacks and emergency cases.In a real scene, we are concerned about the abnormal behavior in video surveillance. When the abnormal behavior is identified in time, video surveillance informs the occurring risk behaviors and types to the monitoring person in the monitoring areas, so that timely detect and prevent more dangerous activities. So how accurate detection and identification of abnormal behaviors are the main contents of this study.The video images which captured by the environment and the electronic device itself will produce blur, incomplete and other adverse effects. In order to improve the image data, inhibit unwanted distortion and enhanced some important image features for the subsequent treatment, reduce the amount and complexity of the calculation of follow-up treatment, so in this paper video images pre-process, including the gray level transformation, threshold segmentation and median filter.In the moving target detection, this paper combines the background subtraction and temporal difference. In a real scene, in the accurate detection of moving targets, the shadow of an object has a greater impact for accurately identifying targets, so this paper eliminates the shadow processing. This paper introduces the HSV space shadow detection and the RGB space shadow detection. Meanwhile, the author concludes that RGB space for shadow detection is a more suitable for detection of moving targets after comparing and analyzing the two methods on eliminating the shadow.In recognition of abnormal behavior, this paper uses the characteristics of abnormal behavior. A couple of criterions are given to analyze whether people disappear, climb, fall or something left in monitoring areas and alarm. Experiments demonstrate that the approach which is proposed on abnormal behavior recognition is easy, fast and effective and achieve good recognition effect.