Research On Object Tracking Algorithm Based On SIFT Feature-points Matching

Moving Target detection and tracking is a hot research topic in the domain of computer vision, which is widely used in real live, such as Bank Management, Medical Research, traffic management and so on. In this paper, a method based on feature point matching of SIFT Algorithm is used to achieve the target tracking. SIFT algorithm has good invariance against scaling transform, image scaling and rotating. There are four sections composed this article, Main work are as follows:(1) The part of the moving target detection, For the background site, amending the background by a learning rate, using morphological processing to make foreground objects into a connected domain. This dissertation used a method of two scanning based on connected domain this method can effectively separate connected domain accurately.(2) Using SIFT algorithm to match feature points between targets segmented, write down the number of matching points. The two of matching the most number of feature points is the same goal.(3) The size of segmented target is much smaller than the size of the original picture. Therefore, some improvements has been made for SIFT algorithm. We adjusted the number of down-sampling of a picture and Gaussian pyramid layers, not only guarantees the number of matching points between the same target,false match rate is very small, but also shorten the time of the match. The experiment proved that the efficiency of time of the algorithm is improved.