Dynamic Image Object Identification And Tracking Research

Dynamic image processing technology, mainly includes moving object detection, classification identification and tracking, understanding and description some behavior of moving target. Target motion detection and tracking technology has been used widely. It has been a research direction to domestic and international numerous researchers for many years However, because the complexity of video sequence, so far no one method can be used in various venues comparatively perfect. Therefore, it is still a huge challenge to studying a kind of efficient and robust method of target tracking in the field.This paper first introduces the basic knowledge of digital images, including noise smoothing algorithm, neighborhood average method and median filtering, corrosion and expansion of mathematical morphology method and choosing threshold method (the categories of variance method), this paper mainly Comparative analyzes two improved weighted median filtering algorithm(weighted adaptive median filtering and weighted fast median filtering), To compare the two new methods get conclusion: weighted adaptive median filtering algorithm has better robustness on de-noising between weighted fast median filtering algorithm. Next, the basic principle and method of the dynamic target recognition are introduced in detail, including adjacent frame differential method, symmetrical finite difference method, background difference method and gaussian mixture model. this paper get advantages and application scope of each algorithm through experiments and put forward improving symmetrical difference method. That is, first do difference operation to three consecutive frames image two and two, then using the improved adaptive median filtering de-noising, and then applied the most kinds of variance method to select threshold to get binary image, finally take logic operation to difference image detection the moving targets. Compared with conventional methods, the improve symmetry difference arithmetic contains more target information, more accurate for the moving object and effectively solved the target empty. In the final, it analysis the basic thoughts of the Mean Shift algorithm and the density estimation theory, and successful application the Shift algorithm to tracking dynamic target. Through the experiment under the MATLAB 7.1 can be seen that the tracking effect is remarkable under the simple background and no shelter by the complex environment.