| In the changeable world, dynamic image provides richer information than that of single image. The function of video moving targets detection and track is to extract moving information of the target. As a hot field of the machine vision and the computer image processing, moving targets detection and track has subjects of artificial intelligence, pattern recognition, automatic control, etc. be involved in and meanwhile finds wide applications in the fields of robot navigation, military guidance, intelligent transportation system, medical image analysis and booming smart family. For the current defects of moving targets detect and track algorithm in complicated application, this study will present several methods to improve the algorithm performance of image tracking.Firstly, this paper will show the introduction to the video moving targets detect and its current development status at home and abroad, the systematic explanation on moving targets detect and track, and the detailed discussion on the image characters and its tracking difficulties in complex backgrounds. For the image character of moving target under complicated background, a target detect algorithm with Horn-Schunck(H-S) optical flow and inter-frame difference method combined will be used to extract the target meanwhile the centroid of the target will be obtained by the projection location algorithm.Secondly, this article will study single target track algorithm. With the utilization of “CURRENT†model and Kalman filter, self-adaptive filtering algorithm is established on the basis of “CURRENT†model, and based on which self-adaptive tracking gate is designed to be used on target tracking. A common nuclear-based tracking method will be discussed and compared with tracking algorithm introduced in this article by experiment. Under different complex backgrounds, moving targets in different scenarios will be tracked by single target tracking algorithm previously mentioned and be analyzed in experiments.Thirdly, this chapter will make discussion on manoeuvrable multiple-target track theory of video images, analysis on current defects of flight path association algorithm. With the information of location and speed as the basic parameters of flight path association and the application of the fight pass association algorithm on the basis of moving parameter prediction, experiment simulation is conducted.Last but not the least, experimental verification and analysis are made based on previous theory and study achievements, the proposed multiple-target and grading tracking method, and the multiple target image sequence shot under complex backgrounds. |
