Moving target detection and tracking is one of the most challenging topics on the computer vision, which involves the image processing, pattern recognition, artificial intelligence and automatic control multiple areas of research. It owns the important theoretical value and wide application in many areas of industrial, medical, military, transportation, etc. Involving multiple scientific fields, although a lot of people keep their researches on this work, this topic still remains many important issues unresolved. Based on a in-depth study of domestic and foreign correlation algorithm, this paper attempts to improve and optimize a classical moving object detection and tracking algorithm.In object detection, this paper introduces the principle and method of Gauss mixture model to detect moving objects, and then improves the detection process of distortion, shadow interference and other issues. By improving the background update rate to improve the detection distortion, the HVS color vector can eliminate the shadow interference. From the results of comparative experiments, through the improvement, the background model can adapt to the interference environment, and do more complete extraction of the moving target to reduce the interference caused by the shadows during the detecting.In target tracking, based on Mean Shift algorithm, this paper focuses on the moving target tracking algorithm. Through the analysis, we find that when the moving object completely occluded or large area disruptors, mean shift algorithm will appear the phenomenon of missing the target. To solve this problem, the Mean Shift algorithm is added to the Kalman algorithm in this paper, which will be cooperated together to track moving targets. It is proved that the improved algorithm can effectively solve the case of lost target after occlusion. |