System Design For Moving Object Detection And Tracking

System for object detection and tracking has extensive application, in brief, it can detect,segment and track the motorial object, and it is the foundation to process the object identification and analysis.This paper designed an integrity system, it consists of three main parts: detect moving object from sequence input stream; segment the object and then track it.On the research of the motion detection, firstly, two main algorithms of motion detection were introduced and advantages and drawbacks of these algorithms were researched. This paper used Adaptive Gaussian Background Modeling to segregate foreground pixels from static background. To responsive the changing background quickly, specifically to solve the "Ghost" issue of Adaptive Background Modeling, this paper provided an improvement algorithmic, by using temporal learning rate of parameter to increases influence of the anterior frames in the modeling. By comparing the results, the improved algorithm is apparently faster in eliminating "Ghost".On the research of the target segment, the usual methods of edge detection are according to gray image. In order to make full use of the color image information, this paper puts forward a new edge detection method of color image based on gradient extreme value. This method is the foundation of using active contour model to fit boundary accurately. This paper put forward the method combined edge extraction with Active contour model to overcome man-made intervene which we should do by using classical Active contour model. It means using edge extraction to detect the edge of the object, afterward, line out the reference points, and then make use of these points as datum marks of Active contour model, consequently, automation segment was implement.On the research of object tracking, we adopt a method combined the Mean-Shift Algorithm based on object feature of color histogram (Camshift Algorithm) with Active contour model. Firstly, use Active contour model finding the accurate contour of the object, afterward, use Camshift orienting the target. We approve this system availability through the contrast experiment, and this arithmetic can find the exact contour, it also can tracking effectively.