Motion Detection Algorithm Research On Intelligent Video Surveillance System

This thesis selected topic from the project for video data intelligent analysis technology research. With the improvement of living conditions and quality, personal and property security attracts more and more people’s attention. The concept of security has occupied a more and more important position in people’s mind which promoted the rapid development of the intelligent video surveillance system. The core function of the intelligent video surveillance system is to analyze the video images which are captured by the cameras. Then the system can detect, recognize and track the things in the image monitoring area.The development of the intelligent video surveillance system and the working principle of the system are analyzed. The intelligent video surveillance system has many advantages compared with traditional video surveillance system because of the intelligent algorithms. Several commonly used motion detection method and the algorithms of background modeling are mentioned. The use of motion detection algorithm and background modeling is very frequent, and their advantages and deficiencies are pointed out.In this thesis, the main research work is as follows:① The adaptive SILTP operator is proposed. The principle of LBP operator which extracts the video image texture feature is introduced, and the limitations of LBP, LTP operators are also analyzed. The shaking leaves, waving water and other complex background can make a bad influence on the motion detection. This thesis proposes an improved SILTP operator to extract the texture information of moving targets, which can eliminate the effects made by the complex background. This operator determines the size of adaptive threshold with the global and local pixel gray level difference which can remove the complex background. The adaptive SILTP operator’s adaptability which can eliminate the impact of the complex background is strong.② The Gaussian mixture model is used for background modeling with the image texture feature extracted by the adaptive SILTP operator. The image texture feature is extracted by the LBP operator, SILTP operator and adaptive SILTP operator. After background modeling, the detection effect of the three kinds of operator through the experiment of motion detection is verified. The result shows that adaptive SILTP operator is better than the other two operators.③ An intelligent video surveillance system is designed and implemented, and the adaptive SILTP operator is used in the system for motion detection and the result is perfect.