Research And Application On Moving Object Tracking Algorithm Based Oncorner Detection

Real- time target tracking is a hot application in the area of computer vis ion, and is also a subject of practical research significance. As the requirements of the modern society for tracking accuracy are becoming higher and higher, corner which regarded as a key feature of the image has advantages in improving the tracking accuracy. Taking it as a breakthrough point, this paper carries out research on the corner detection algorithm and target tracking algorithm, the optimize corner detection algorithm is applied to the tracking system in order to improve the accuracy of tracking.Firstly, based on the comparison of common corner detection algorithms, this paper test the Harris, FAST, SURF corner detection algorithm from the angle of speed and accuracy. The advantages and disadvantages of this three algorithms are analyzed comprehensively in this two aspects. The SURF corner detection algorithm is chosen to apply to the target tracking system. And based on this, this paper optimizes the SURF corner description algorithm, and compares it with the original algorithm.Secondly, after determining the corner detection algorithm, this paper carries out intensive study on motion detection algorithm and motion prediction algorithm, then selects Kalman filter and optical flow method to make testing, comparison, analys is. Based on the feature that Kalman filter can predict the target area accurately and combine corner detection algorithm, this paper puts forward an optimize method of combining SURF corner detection and Kalman filter. In this method, using K alman filter forecast update algorithm is used to update region of interest fastly, in order to narrow the detection range of the corner and improve the speed of the tracking system, so as to meet the requirements of real-time.Thirdly, based on the optimized algorithm, this paper realizes the real- time target tracking system, then the system is used in Volleyball testing to complete the tracking of volleyball, and designs a ranging method of single camera to complete the height ranging of volleyball.Finally, this paper tests and analyzes of real- time target tracking system, completes the testing on the SURF corner detection and the prediction results ofKalman filter, makes tracking performance testing of the real time tracking system by the upthrowing the volleyball.