The Application and Development Trend of Youth Sports Simulation Based on Computer Vision

Based on computer vision technology, this paper presents a human motion analysis and target tracking technology based vision. In terms of moving detection, the current detection is summarized, some experimental results algorithm are given. The background difference method under monocular camera emphatically analyzed. preliminary contour obtained by method. order to obtain smoother contour, mathematical morphology used remove noise, judgment size image connected domain added. A specific threshold set noise block less than threshold. aspect recognition, selects features, including minimum external rectangle ratio, rectangularity, circularity, moment invariant. criteria for selecting features strong resistance obvious distinction. Then, three types images classified recognized. After cross-validation parameter optimization, recognition accuracy significantly improved. show that video sequence collected in field has total 376 frames, frame rate 10 frames/s. Due small traffic, mean shift adaptive feature fusion track every 2-3 frames. And inverse X direction as entering scene id="M2"> out so allowable error distance between 10. weight each dynamically updated similarity candidate model model, which solves problem not robust enough when similar objects occluded interfered achieves more accurate tracking.