Research On Moving Target Tracking Algorithm In Video Sequence

Nowadays,the core content in the area of computer vision is moving target tracking.It has broad application prospects and great practical value.It is a combine of advanced technology and multi-disciplinary research,include automatic control,image processing,artificial intelligence,pattern recognition and computer applications.Moving target tracking is a kind of process that can automatically identifies a target and determines the location of target,then automatically tracks the target.In area of computer vision,it is one of the hot research issues,and it currently widely used in intelligent control,intelligent transportation,automotive driver assistance,precision-guided and other civilian and military fields.Therefore,the image sequence target detection and tracking has a very important practical significance.The image signal information loss during collection process and the complex environments makes target tracking a challenging task.Especially in the tracking process,the occlusion between targets and variable number of targets are the key factors that limits the result of tracking algorithm.This paper' research is mainly about moving target detection,target tracking,data association algorithm of multiple target tracking,occlusion treatment and other important research issues and difficulties in video sequence target tracking.Some of the research methods will be used in the relative application fields,and provide application guidance and theoretical support for moving target detection and tracking.Through the introduction of a variety of technologies methods and key theories,it leads to the main work of this paper,which mainly includes three parts:(1)Target detection:We first introduce some common moving object detection algorithm and the improved visual background modeling algorithm and used this method detect moving target,then segment the multiple targets' motion regions.In the end,the appearance of the target model represent the region area by the bounding rectangle.(2)Target tracking algorithm:we introduced several commonly used tracking algorithms and theory,and summarizes the research about particle filter improvements.We elaborated on the basis of an improved paper made traditional particle filter algorithm and through the experimental comparison we illustrates the improved result of tracking when there is occlusion between targets.(3)Multi-target tracking system:first we introduced multi-target tracking theory,and described the multi-target tracking system architecture.Then we introduced the key step of multi-target tracking-data association in detail,and describes the data association method based on global nearest neighbor algorithm.Finally,the multi-target tracking system experiments illustrated the tracking results.