Study Of Moving Object Detection And Tracking Algorithm In Complicated Scene

Intelligent Video Surveillance is to use the methods of digital image Processing and computer vision to analyze real-timely and automatically to locate, recognize and tracking the Target by using vidicon to record Picture Series, then we can analyze and judge the target's actions. So the study of moving target detection and tracking technology which act as the key technology of Intelligent Video Surveillance is especially important. Magnitude researches on motion detection have been done by many experts all around the world, but almost all the old researches only focus on the simple and static background scene. Recent years, along with the wide-range application of the Intelligent Video Surveillance in daily-life, the moving target detecting and tracking technology develops fast and becomes more and more popular.This Paper mainly discusses the fundamental theories and key technologies of moving targets detection and tracking. The following topics are researched in details, such as detection and extraction of moving objects, especially the problem of moving objects detection in dynamic scene, carries out the moving objects detection algorithm based on adaptive background model. We also present a background modeling algorithm based on Non-parametric Kernel Density Estimation (NKDE). This algorithm consider the dependence both in temporal and space domain, and modeling both the background and foreground, when the traditional algorithms always only model the background just using the temporal dependence. Besides, the presented algorithm utilizes the non-parametric model to estimate the density function, overcome the problem that the single or multiple gaussian model can't approach the density function adequately. Thereby, we solve the problem of moving objects detection in complicated scene, such as camera dithering, branch waving, ripple .etc periodic motion and rain or snow weather scene. Experiment results shows that the presented method has very well effect, runs fast and robust.This Paper also studies the key technology of moving object tracking, especially stress on Kalman filter, and carries out a moving object tracking algorithm based on Kalman filter. It completes the tracking by object detecting, motion estimation, object matching and tracking. We obtain the moving object using the previous detecting algorithm, and describe the shape character (such as object location, external rectangle size .etc); Then the Kalman filter is used to estimate the object location in next frame and confirm the search range; finally object matching is used to find the object's corresponding relation between each frames in video sequence, and track the moving object.