Research On Moving Object Detection And Tracking Algorithm In Video Surveillance

In recent years, computer vision and image processing is a popular research direction in the computer science. As one of the hot topics belong to the discipline, moving object detection and tracking technology in the video surveillance is being researched by many scholars, but there are many problems unresolved yet. For example, detected object is incomplete, object tracking is failed in the complex background and so on. 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 in video surveillance, and improves the shortcomings of the existing algorithms.In this paper, not only the application background of the moving object detection algorithms and tracking algorithms in the video surveillance, but also the domestic and foreign researches development of this subject are introduced first. At the same time, some relevant theoretical knowledge used in image processing such as gradation processing of images, binarization processing of images, color space transformation of images and mathematical morphological processing of images are researched, which provides a theoretical basis for improving related algorithms.In the moving object detection, firstly,several common typical moving object detection algorithms are researched, then their advantages and disadvantages are contrasted and analyzed through experiments. Secondly, the traditional three-frame difference moving object detection is improved for its shortcomings that detected object is imcomplete and has internal empty. In this paper, the improved algorithm uses local adaptive threshold for binarization processing of difference image, instead of using a fixed threshold. At the same time, the contour of binary images are filled, which improves the shortcomings of the traditional three-frame difference algorithm and makes detected object more accurately and fills the empty in the object.In moving object tracking, firstly, the classification of the current moving object tracking algorithms is introduced. Then this paper focus on the popular algorithm in the current field(Camshift algorithm). However, when interference occurs on the color between the target and the surrounding background, or the background changes is relatively intricately, the traditional Camshift algorithm may appears failed tracking. For this shortcoming, this paper improved the traditional algorithm based on feature point matching algorithm of SURF with RANSAC purification. When it is judged that the object loses, the SURF algorithm are introduced to reposition the object area in order to find the missing object to achieve the sustained tracking.