Research On Target Detection Algorithm In Intelligent Monitoring System

Public safety is the issue that being paid much attention to. With the rapid development of information technology and computer vision technology, intelligent video surveillance system has become one of the critical technology in the field of public security, as well as the hot topic among scholars. In the intelligent video surveillance system, the primary task is to extract the moving object, which is called target detection. Moving target detection is the key to the system. The following work are target tracking, target recognition, abnormal behavior recognition. Whether the moving object is accurately extracted will directly affect the following work, and it will also affect the property of the entire surveillance system. So, how to extract the moving object from the complicated monitoring scenes real-time and accurately is the key to intelligent monitoring system. This paper mainly researches the target detection technology in intelligent monitoring system, and focuses on background subtraction algorithm, the Self-Organized Background Subtraction(SOBS) algorithm, the improved SOBS algorithm and Gaussian Mixture Model(GMM). Moreover, it researches several intelligent video analysis applications, including intrusion detection, and abandoned objects detection. This paper lays a good foundation for the intelligent video surveillance system. The main job and contributions of this paper are as follows:1. The three development stages of video monitoring system is researched: ordinary artificial monitoring system, semi-automatic control system and full automatic intelligent monitoring system. Through introducing the three systems, their shortcomings are compared, then ordinary artificial monitoring system and semi-automatic monitoring system are improved and perfected to produce automatic intelligent monitoring system which is currently widely used.2. Three current main methods of moving object detection techniques are researched, including frame difference, optical flow and background subtraction. Principles of typical algorithms among these methods are introduced in detail, and the advantages and disadvantages of each algorithm and their scope of applications are analyzed experimentally. After considering many factors, background modeling method is selected as our main object detection algorithm.3. The research mainly focuses on background modeling method based on Self-Organized Background Subtraction(SOBS) algorithm. SOBS algorithm is presented in three parts: the background model, foreground detection and background updating. The advantages and disadvantages of SOBS algorithm are analyzed, and the improved SOBS algorithm is put forward. Through contrast experiments, the results show the validity of the improved SOBS algorithm.4. The moving target detection technology which applied in intelligent monitoring system is explored. Then, combines improved SOBS target detection and Kalman Filter target tracking technology, the paper researches application modules such as intrusion detection of sensitive area and abandoned object detection. Tests results show that these application modules are effectiveness.