A Study Of The Motion-Tracking Algorithm Based On Scale-invariant Optical Flow

With the popularity of digital cameras and the rapid development of computer graphics video processing technology, the application of computer vision is more and more widely. Video sequence motion detection and tracking technology are one of the important research directions in computer vision applications. This paper mainly studies on moving target detection and tracking algorithm in video sequences.For the case of there may be blocked or self-occlusion in the process of moving object tracking and change of target shape, a new algorithm of moving target detection and tracking which combined with scale invariant feature optical flow and dynamic update policy objectives was raised. The main work is done as follows:(1) Moving target detection. The primary task of moving object detection is to extract moving object from the background region, so as to achieve the purpose of object detection. In this paper, background subtraction method to detect motion from a sequence of images to track objects. In the video image by solving video frame and background model difference detect moving targets, and by morphological filtering technology to remove noise moving target area, and given by a rectangular frame, to achieve object detection, and then fixed by rectangular box.(2) Moving target tracking. Moving Moving target tracking process is to identify the same object in different frames of the image sequence position. This paper presents a new dimension moving target tracking algorithm for optical flow and constant update dynamic template using SIFT features rotation invariant and scale invariant characterization moving objects, and using the pyramid method to calculate optical flow feature point vector, achieve the purpose of tracking, dynamic approach to problem solving template update tracking target occluded track failed.(3) Realize the moving target detection and tracking algorithms C++ and Open Cv library were used to program. From an implementation point of view of the effect detection and tracking, the algorithm is robust and effective for the case of moving target was blocked or target morphological changes.