Moving Object Tracking Based On Stereo Vision

Moving object tracking in video sequences is a process of tracking processing based on images captured by video camera.It has been widely used in secure monitoring,space technology,medical and healthy care and other fields,which means it is closely related to our life.Therefore,the research of moving object tracking becomes one of the research hotspots.People have done a lot of research on this issue in recent years,and some practical tracking algorithms are proposed during the period.But there is still a long way to go to propose a practical algorithm which has good robustness and strong applicability.So the subject is still very challenging.Traditional methods of moving object tracking are based on two-dimensional spatial information.But when dynamic occlusions occurred between moving multi-targets,we can not distinguish moving targets using traditional methods,which means tracking failure.Because of the depth information in three-dimensional space,three-dimensional coordinates of moving targets would not change suddenly before and after occlusions.That's why three-dimensional coordinates are more suitable in complicated and changeable situation between moving multi-targets.Specifically,the main research contents of this paper are as follows:1.The background subtraction method of merging-based disparity map based on color space transformation is proposed.This paper introduces some traditional methods of moving objects tracking as well as their respective characteristics.We also carry out some simulation experiments,and the results show that,in view of the fact that we would extract the feature points of moving targets in the following process of this paper,the algorithm of moving object detection based on background subtraction method is more effective.Owing to shortcomings of traditional background subtraction method,which includes the sensitivity to illumination changes and the shadow produced by light irradiation,the effect of traditional method in dealing with these issues is not obvious.To solve the problem,firstly this paper realizes the conversion of images captured by video camera with color space model.On this basis,this paper conducts the background subtraction method of merging-based disparity map.Then this paper carries out morphological operation on obtained binary images.Results of simulation experiments demonstrate that the proposed algorithm is more effective than traditional methods,which lays the foundation for the next step.2.A moving object tracking method based on three-dimensional coordinates is proposed.Firstly,we extract feature from moving objects with sift algorithm.Then we carry out feature points matching extracted from moving targets.And we cluster centers of feature points reconstructed by three-dimensional coordinates with FCM algorithm.Finally,we track the three-dimensional hub of targets clustering with Kalman filtering.And the process of combining three-dimensional coordinates of moving targets with Kalman filtering is deduced in detail.Results of simulation experiments demonstrate that the proposed method is more adaptable and more effective than traditional methods based on two-dimensional coordinates.The proposed method is feasible and effective.