Study On The Algorithms Of Moving Object Detection And Tracking Based On Video Image

The moving target detection and tracking on video image is a key technique and the major research direction in the field of computer vision, It has been widely used in the intelligent transportation, visual navigation, safety monitoring and many other fields. Moving target detection and tracking are two inseparable parts in the accurate target tracking process. Its basic task is to realize real-time detecting of moving targets emerged from video sequence frames, to obtain the target parameters, to match and track moving targets, and then to get the trajectory of the target. This peper mainly studies the system structure of the moving target detection and tracking on video image, the video image processing technology used in the system, the moving target detection and tracking algorithm, and realizes a moving target detection and tracking system.First, an background difference algorithm based on the cumulative difference to update the background is proposed by analyzing of the commonly used moving target detection algorithms and their advantages and disadvantages. The algorithm based on the accumulative difference concept, makes improvement on the adaptive background modeling method, which can effectively eliminate the interference of noise influence, scene changes and so on. With regard to the threshold judgment of target extraction, this paper puts forward a kind partition method of gray histogram threshold based on the difference between the two main peaks.Then, the CAMShift algorithm and Kalman filter algorithm are analyzed and compared, and a kind of moving target tracking algorithm based on the geometric characteristics is proposed which combines the Kalman filter and histogram matching. The target centroid and the length and width of circumscribed rectangle are used as the geometric characteristic parameters of object. After using respectively two groups of Kalman filter to predict, target matching is made according to the similarity of target gray template.Finally, the moving target detection and tracking system based on video image is realized in the VC++6.0integrated development environment, using the MFC application framework and OpenCV computer vision library code. For the input video file, the system can real-time detect moving objects in the video and tracking.