Research On Object Tracking Based On NMF Coding

Moving target tracking in video sequences,It has been one of the core research topics in computer vision field.Tracking algorithm has been developed for a long time and because of the people's in-depth study,It has been widely used in many fields such as video surveillance, face recognition, medical image analysis, robot navigation,security monitoring and so on.Object tracking is to find the location of the target in each frame of the image, and mark them out. However, the robustness and accuracy of the tracking algorithm is still a challenging problem. The appearance change of the target or background, occlusion of target, the sudden change of light, camera shake and so on will have a impact on the tracking results. These are places where we need to keep improving.This paper analyzes and studies the common challenge problems in target tracking.The main research contents and results are as follows:1.A vedio tracking method based on manifold-constrained non-negative coding(MCNC) is proposed.The method is based on non negative coding,and the manifold geometry of the local feature space from neighboring superpixels and the distance between feature vectors and dictionary is incorporated,make the dictionary can be more accurate for encoding,improve the accuracy and stability of encoding.This method can effectively solve the problem of background similarity, occlusion, object deformation, etc., and improve the robustness of tracking algorithm.2.A form of encoding visual tracking method based on dual dictionary is proposed.This algorithm describes a novel construction method of codebook,different approaches to the traditional K-means clustering approach, it is to construct a positive and negative dictionary by extracting the features of the square block and the target around the target.Through our experiments can be seen that this method can effectively deal with the common problems in visual tracking such as serious occlusion, illumination, fast motion and so on, and can get good tracking results.