Visual Tracking Based On Position Prediction And Double Matching Method

Object tracking is an important research direction in computer vision, with the purpose of realizing accurate object position in every frame, so that we can analyse behavior and form change of moving object in next step. It is wildly applied in multiple situations, such as modern military, traffic navigation, security surveillance and medical image sequence processing. Now, there have been quite a lot of excellent algorithm. However, due to multiple challenging factors such as background interference, illumination changing and similar object, how to design a robust object tracking algorithm is still worth researching.On the basis of existing target tracking methods, in this paper, we design a new tracking method that build search model, determine search center and search radius all according to the motion laws of the target. In order to improve the possibility of catching the right target, this method make the search center and target location closer by modeling moving target and make search range cover target location more effectively by adjusting search radius adaptively.For computing the match degree of target template and candidate area, we combine sparse representation method and gray value error judgment method together, and analyze image match errors of these two methods. Then confirm the target location by weighted targets’coordinates. Our proposed method avoid missing targets in some video frames effectively and combine two positioning methods together to make sure that we can find the target location more precisely.The experimental results show that our algorithm can stable track object stably in video sequences in the treatment of the object shelter, complicated background, and illumination changing. Compared with the current popular algorithm, our algorithm have the superiority in tracking accuracy.