Font Size: a A A

Research On The Key Technology Of Moving Object Detection In Image Sequence Under Natural Environments

Posted on:2004-08-04Degree:MasterType:Thesis
Country:ChinaCandidate:Z W LiFull Text:PDF
GTID:2168360152457040Subject:Circuits and Systems
Abstract/Summary:PDF Full Text Request
Along with the development of computer hardware and computer vision technology, the applications of vision images have been paid more and more attention to. Moving object extraction is one of the first and most important steps in image sequence processing. Moving shadow detection has been neglected in many popular methods of moving object detection these days. Although shadow detection has been done in some methods, it hasn't been studied in depth. Existing results can't meet the demands of application in practice. In particular, moving shadows of objects can affect the correct localization, measurements and detection of moving objects. Therefore, shadow detection is a key technique in image sequence applications, which needs a further study.Based on the analyses of image models, this paper has generalized shadow character of gray and color image sequence. Particularly, the intrinsic character of shadow in color image sequence that it doesn't change the color of object surface has been pointed out. On the basis of two kinds of image models in color space, shadows of objects have been detected by background subtraction approach. According to shadow characteristics under image illumination model, a novel method that takes normalized cross correlation coefficient as metric function has been proposed to detect moving object with self-shadow.Finally, with VisualC++6.0 the method proposed in this paper has been simulated. On the basis of dynamic and static background, moving object detection has been realized in both gray and color image sequence. Experimental results have showed that the method can be used in complicated natural scene and detect object effectively. Also, it is of some robustness.The methods proposed in this paper have good application prospects in vehicles tracking and counting as well as target recognition that takes contour and area as the main reference characteristics.
Keywords/Search Tags:Image sequence, Dynamic background, Moving shadow detection, Moving object detection, Normalized cross correlation function
PDF Full Text Request
Related items