Research On Target Tracking In Panoramic Vision

In recent years,because of its 360-degree viewing angle,panoramic vision has great potential for development in areas such as traffic control,video surveillance,and unmanned driving.However,due to the non-linear geometric distortion of panoramic images,there are Due to the similar background,occlusion,rotation and other factors,the traditional particle filter algorithm can not be applied to the panoramic camera normally.Therefore,this paper studies how to effectively target tracking in panoramic vision.For the problem of tracking failure caused by target distortion in panoramic images,a method is proposed to solve this problem by mapping the coordinates of the panoramic image to the unit spherical coordinate system.Because the target in the panoramic image is geometrically distorted during imaging,resulting in nonlinear resolution,which makes it impossible to use the conventional method of calculating the feature map for target matching,a unit spherical mapping method is proposed to map the geometric distortion on the panoramic image.On the unit sphere,the spherical coordinates are used for sampling.In doing so,the sampling of the target conforms to the geometric characteristics of the image.At the same time,the size of the sampling frame is changed according to the spherical coordinate characteristics.Aiming at the problem of target tracking failure caused by the background during the target tracking process,a color feature and shape feature fusion method is proposed.In the ordinary video tracking process,if the target color is the same or similar to the target background color,tracking failure often occurs.In this paper,color histograms are used to represent target color features,Chamfer Match is used to represent target shape features,and finally Bayesian is used.The fusion recursion formula expresses the overall characteristics of the target and improves tracking robustness.For the problem of tracking failure due to the change of target scale in the target tracking process,an adaptive target region extraction method is proposed to solve this problem.Firstly,the target contour segment in this scene is obtained through the unit sphere mapping model.After that,the parameters of the target trapezoidal frame in this scene are obtained through analysis,which makes it possible to accurately represent the target contour,reduce the background noise in the tracking process or the incompleteness of the target area,and improve the target tracking accuracy.Finally,the experimental results show that the proposed method can adaptively adjust the size of the trapezoidal frame and effectively solve the target tracking process,which has a good effect due to similar background,occlusion,rotation,etc.,Effectively improves the robustness of panoramic visual tracking.