Moving Object Detection And Tracking Algorithm Based On Video Sequence

In recent years, computer vision and image processing are popular research directions in the computer science. As one of the hot topics belong to the discipline, moving object detection and tracking of the video images sequence is being researched by many scholars,but there are many questions unresolved yet. So there is important significance to keep on researching this subject. This paper focuses on the research of moving object detection algorithms and tracking algorithms of video images sequence, and improves the common algorithms of the discipline correspondingly based on its shortcomings.In this thesis, not only the application background of the subject, but also the domestic and foreign researches development of moving target detection and moving target tracking based on video images sequence are introduced firstly. At the same time, some relevant theoretical knowledge used in image processing such as grey level transformation, binary image, edge detection and mathematical morphological processing are researched. Then several common traditional moving target detection algorithms is introduced, which provides a theoretical basis for improving related algorithms.In the moving object detection, two kinds of algorithms of moving object detection are proposed according to the scope of application. Firstly, since the traditional three-frame difference algorithm is sensitive to noise and prone to form internal "holes", an improved algorithm named three-frame difference algorithm of rated thresholds is proposed combined with LK optical flow matches. In this way, the threshold segmentation of the traditional three-frame difference algorithm converts into another mode which combines the threshold segmentation with the region segmentation. At the same time, the corners calculated by the optical flow method before are used to improve the contours of the targets. This algorithm can get better detection results than the original one. Then, for the ViBe algorithm, a kind of background model algorithm proposed in recent years, it is easy to form the "smear"phenomenon and is sensitive to light mutation. This paper takes advantage of the edge detection, three-frame difference and other components to form a detection module, making it adaptively correct the erroneous detection prospects. This Improved algorithm improves the shortcomings of the original algorithm to detect the correct prospects of target.In moving object tracking,Camshift is usually widely used in the field of object tracking in recent years. However, when interference occurs on the color between the target and the surrounding background,or the background changes is relatively frequently and fiercely,the traditional Camshift algorithm may lose effect. This paper proposes an improved Camshift algorithm based on the feature point matching algorithm of ORB. When the phenomenon that the target may lose occurs, the introduced ORB algorithm for feature point matching can find the missing target. In this way, the sustained tracking can be achieved.