Target Tracking Method Based On Adaptive Frame And Motion Model

In the real environment,moving targets tracking in complex scenes have a series of problems,such as deformation,occlusion,rapid movement,and scale changes,which poses a higher challenge to the accurate tracking of moving targets.Aiming at the problem that in the process of a moving target tracking,the aspect ratio of the target in the image will change due to the changes in the obser vation angle and posture of the target relative to the camera,this paper designs two independent filters to estimate the target’s length and width.Verification and comparison with existing tracking methods on public data sets show that this method can effectively improve the adaptive ability of aspect ratio scale while also meeting real-time requirements.In the shooting range experiment,when a target with a large aspect ratio such as a rocket or a missile rotates,the tracking will fail.Aiming at the problem,this paper takes the smallest bounding rectangle with an inclination angle as the initial state of the target,and designs an independent Angle filter to estimate the angle change when the target moves.Experiments show that this method solves the problem that the correlation filtering method cannot effectively track targets with large aspect ratios.When the moving platform observes the moving target,there are problems of target occlusion and rapid background movement due to the movement of the ca mera.Based on correlation filters,this paper uses optical flow method to achieve image registration to eliminate background motion,map the historical position of the target to the same image,and establish a motion model of the target on the image to pr edict the target in the current frame.A long-term filter is designed to store pure target samples and determine whether the target is occluded;When the target is occluded,switch to the long-term filter to search for the target in the candidate area according to the position predicted by the motion model.If the maximum response value reaches the set threshold,the target is successful recaptured.The comparison test with the existing algorithm on the public data set and the real photo data set of the photoelectric pod shows that the method can effectively solve the problems of target occlusion and rapid background movement due to camera movement.Aiming at the problem of less background texture and difficulty in image registration when tracking a moving target,combined with the spatial positioning requirements of the target,based on the long-term and short-term filter,this paper proposes to predict the target by fitting the three-dimensional spatial motion trajectory of the target,and project the predicted three-dimensional space position to the image pixel coordinate system to predict the target loaction on the image.A comparison test with existing methods on the optoelectronic pod data set shows that this method can effectively improve the accuracy and success rate of target tracking in the optoelectronic pod.