Research On Moving Objects Recgonition In Complex Backgrounds

Detection and tracking of moving objects in image sequence is an important area in the domain of digital image processing, pattern recognition and computer vision. Its applications include that robot navigation, security surveillance, medial image analysid and virtual reality.This paper mainly discussed the topics as follow: detection and extraction of moving peoples with complex background, and tracking of moving objects and so on. Concerning the object detection, discussed the detection method based on the complex background in detail, researched method principle of the background initialization the establishment and the update of background model. Improved the background update algorithm based on the IIR filter, the paper proposes a new method of background update-coefficent obtaining, which can update the background adaptively in different video sequences. The experimental result show that the method has the feasibility when considers the validity and the complex degree of the algorithm.On the research on the extraction of moving objects, in order to segregate the targets from the foreground pixels compeletely and obtain the characteristics' description of moving targets ,based on the segregation of the differ-image and according to space connectivity and area characteristic, the paper removes the noise by used the morphology filtering and connectivity detection. After extracting the complete object area, establishes the vectors of the object characteristic for following object tracking. The experiments' result show that the method can extract the moving objects.The statistics proportion of the object extraction also show the validity of the method.Based on Camshift algorithm and Meanshift algorithm, this paper judg the shelter through Bhattacharyya parameter and then select Kalman filter or condensation algorithm tacks the objects. By the experiment, it obtains some valuable conclusions.At the end of the paper, a small-scaled motion analysis system is built under the Windows platform based on OpenCV library, which confirms the validity of algorithms in the paper.