Moving Objects Detection System Of Variant Backgrounds

Segmentation meaning moving objects as to a video image sequence by a certain method is the intention of the moving object detection. Moving object detection is important in a large number of fields. For example, segmentation moving objects is the first step of the objects recognition in the computer vision field; through video segmentation, compress efficiency can be improved in the image compress coding. Recent years, a coding method based on the content has been brought forward because of the advance of MPEG2&4, different coding with the different content, firstly is to segment video for some semantic regions through different object detection method.Recent years, many experts and researchers of the world suggested many effective segmentation methods. But there is not have a method, that is common and available to all kinds videos at present, that is video content, peculiarity itself is too complex and influenced by the noise.In this paper, we design a codebook-based background modeling method to detect moving objects from video sequences. The concept of code vector in codebook originated from video compression technology is used to describe the background characteristic of different pixels in terms of feature parameter. Code vector describe the rule of each pixel variation distributing in this videos. Because a code vector corresponds to a model, so, the codebook of the pixel can full enough to describe the pixel's variation in such section.As to a video sequence which is belong to the same scene, there are some correlations for the pixel with same position in different frames, thus, the code vector corresponds to the pixel can classify according to clustering judgment of background model, the numbers of code vectors of codebook for each pixel after classified are different, so the background description is more flexible and abundant, and the background codebook include not only the unchanged stationary background, but also the dynamic background in the long sequence video.The test results show that this algorithm can handle scenes that contain moving background or illumination variation, and it achieves robust detection for compressed videos, and it can validly solve problem that is intermixed by changed background and moving objects(trees swaying). This algorithm improves the quality of detection and makes the applicable range wider.