Theory And Application Research On Soft Feature Object Tracking

Image moving object tracking,as one of the hottest research topics in the international field of photogrammetry and computer vision,has achieved extraordinary progress after more than 50 years’ research and development.According to the research on the history of the past and the present situation,image object tracking has been researched for a long time and has made great progress,but object tracking in complex background is still not well resolved,the main reason is the existence of the deformation,occlusion and small amount of 3D image information,the moving object image has the nonlinear change rules of continuous and smooth,therefore,to carry out real-time,stable and accurate tracking of moving object,the nonlinear change feature need to be analyzed,extracted and modelled by a new set of theory and method,so the establishment of effective linkages between nonlinear change feature and moving object will give the image object tracking a great help in complex background.A novel theory to object tracking based on the soft feature which is including basic theory of soft feature,horizontal soft feature,vertical soft feature and three dimensional soft feature is proposed in this paper to overcome the problem of deformation,occlusion and small amount of 3D image information.The following is the general idea.Firstly,in the direction of the biological "visual elements" to work together,a novel approach of horizontal soft feature tracking method based on visual quantum is proposed,which takes full advantage of visual invariance of quantum frequency step to achieve "biological vision" horizontal perspective tracking.Then,A novel approach to object predictive tracking based on the vertical soft feature is proposed in this paper to take advantage of the same frequency change differential in foreground,In order to achieve the precursor prediction tracking by extracting the target band information and the establishment of soft feature constraint model.Finally,a novel approach of 3-D soft feature object recognition and tracking method based on curved space field is proposed,which takes advantage of constraints on the spatial distribution.A large number of experiments based on the international standard database and the actual application test show that the proposed approach which combine horizontal soft feature,vertical soft feature and three dimensional soft feature,has overcome the illumination changes,scale extension,occlusion,small amount of 3D image information,and achieves good tracking results.Compared with the state of the art tracking method,soft feature tracker has the following advantages:(1)The proposed approach of horizontal soft feature based on vision quantum takes advantage of visual invariance of quantum frequency step.Step invariant features has independent and binding description using multiple visual quantum to solve the problem of shape change,scale change and many variable structure factors impact on the moving target tracking.(2)It has a better anti interference on the target full occlusion by using the vertical soft features and constraint model tracking the moving target,so that this method has high accuracy,stability and better robust performance.It’s a better solution to solve the problem of object lost under full occlusion by using the precursor shock strength to predict the target precursor.(3)Three dimensional soft feature tracking method using curved space field to describe 3-D object,this method not only preserves the speed advantage of 2-D object recognition,but also integrates the part 3-D information of 3-D object recognition,so it shows a higher real-time tracking performance.