Study Of Moving Object Tracking Technology Based On Computer Vision

Moving target tracking technology gradually becomes the hot topics in the study of computer vision and image processing,because of it has a wide range of applications in many fields,such as military guidance,security surveillance,visual navigation,,intelligent transportation,video encoding,medical diagnosis,analysis of meteorological and astronomical observations and so on.However,the task of robust tracking is a focus concerned by researchers that regarding occlusion,object deformation,illumination variation,real-time restriction,etc.It is also an urgent problem in practical application.This paper mainly focuses on the target tracking algorithm and the swivel table control in the target recognition control system with compound axis and dual axis.For the target Tracking algorithm,because of the TLD(Tracking-Learning-Study)algorithm shows good tracking performance in tracking a moving target for a long time,so this paper studies it deeply.But it also has some disadvantages,such as slowly processing speed,not highly tracking accuracy,vulnerable to interference.For this disadvantages,a improvement algorithm was put forward in this paper.The improvement algorithm is mainly using KCF(Kernelized Correlation Filter)tracking algorithm to improve the tracker of TLD algorithm,and then combining with the detector and the study module to form a new algorithm.Finally,through the experimental results and error analysis,we can know the improvement algorithm has a significant improvement in processing speed and tracking accuracy.For the control system,this paper mainly studied its working principle and the control information to the swivel table.The system reduces the error by combining macro-control and micro-control,and achieves accurate tracking.But it takes some time in the target tracking process,and it also has a delay when the control information transmits to the swivel table,and it will take some noise when the information transfer,so the tracking accuracy will be reduced.For these programs,this paper used the Kalman predictor to predict and filter the control information,and finally through experiment to verify the effectiveness of the method.