Research On Moving Target Detection Based On Image Compensation Under Rotating Camera

With the advent of the era of artificial intelligence,computer vision,as an auxiliary tool for people's daily life,has become more and more important.Object detection is the branch that is most closely related to people's daily life in computer vision.In recent years,it has been more and more widely used in various fields.At present,the research of target detection is mainly focused on the static platform,that is,the camera is in a stationary condition.However,in practical applications,target scene detection needs to be carried out under the moving platform or camera movement conditions.At this time,the method for stationary situations will not be applicable.Therefore,the moving target detection technology under the moving platform needs to be studied.In the image collected by the camera rotation,in addition to the movement of the foreground,it will also be accompanied by the movement of the background.Due to the complex imaging mechanism under the horizontal rotation of the camera,these background motions are not linear motions in the traditional sense.If the related method under the static condition of the camera is used,a large error will occur and the foreground moving target cannot be detected.In this paper,aiming at these problems in the process of camera rotation,this paper designs and implements a moving target detection method based on image compensation and corner optical flow tracking during camera rotation.The specific contents of the method are as follows:(1)Based on the imaging mechanism of the camera,dynamically model the imaging process of the camera under horizontal rotation.In the established model,the matched feature point pairs in the previous and next frames are used as input to solve the parameters of the motion model.Based on this dynamic model,the image is compensated to solve the problem of non-linear motion caused by camera rotation.The preliminary processing of the image lays the foundation for subsequent target detection.(2)In order to further eliminate the false detection of foreground moving targets,this article establishes a common coordinate system based on the principle of cylindrical projection,in which the optical flow tracking method of moving targets is used to eliminate the false detection of foreground moving targets.In order to reduce the effect of mismatching of background corner points during the continuous detection of moving targets,this paper screens corner points based on point tracking technology and the principle of epipolar constraint,removes the mismatched background corner points,and keeps only foreground moving target corner points.