Font Size: a A A

Moving Target Tracking Algorithm Based On Dynamic And Static Features Integration

Posted on:2017-02-05Degree:MasterType:Thesis
Country:ChinaCandidate:Z W SunFull Text:PDF
GTID:2348330488990752Subject:Computer technology
Abstract/Summary:PDF Full Text Request
Feature selection in multi-target tracking is an important factor of tracking stability,and it is a hot issue of feature tracking research.In this paper,a new method which is based on dynamic and static features has been proposed.In this way,we can respectively obtain dynamic optical flow features and inside of the static local edge gray(gray)of moving targets and gradient direction features(edge).Finally,the characteristics of the target model will be established in choosing flexible block way,which has achieved the moving target tracking algorithm based on dynamic and static features integration.In this paper,the main points are as follows:(1)The Gaussian mixture model(GMM)will be used to obtain motion areas.Then,detection area will be processed which combined with morphological operations and get the complete detection target subsequently.(2)We studied Lucas-Kanade optical flow which used the forward-backward optical flow Nearest Neighbor algorithm.Sobel edge forward-backward match error method is proposed on the basis of previous algorithm.Then,combined with the consistency of the direction of movement,the dynamic features model of moving targets will be established.(3)The local static characteristic model of the moving object is established by the method of the flexible block mode and the first order edge gradient direction and the zero order along the direction of the inside of the gray gradient feature(G&E).(4)A feature matching model based on the combination of dynamic features and static features is proposed.For the sub block with dynamic feature matching as the effective sub block,the static features of the effective sub block is the final matching feature.Then,the dynamic and static features model is established.(5)Based on feature model,a multi-target tracking management scheme and tracking and locating matching mechanism are established to realize multi-target tracking.In the target tracking,it is not stable to track the target by using a single static feature or just using the dynamic information.Combined with the research status of the tracking,feature model in combination with dynamic and static features is proposed,which achieved tracking by establishing multi-target tracking algorithm framework.Finally,the PETS data and self-collected data were be used to test.Experimental results show that the proposed algorithm is feasible in moving target occlusion.
Keywords/Search Tags:Multi-target Tracking, Sobel Edge, Forward-Backward Error, G&E Features, Flexible Block
PDF Full Text Request
Related items