Contour-based Pose Estimation For Remote Space Target

The three-dimensional pose is an important external parameter to describe the space target,and it is widely used in space target recognition,vision navigation and rendezvous and docking.Optical measurement is the contactless full measurement.On account of the strong anti-interference,simple equipment,and robust performance,it is widely used in space target pose estimation.The space target pose estimation based on monocular vision optical measurement has become a research hotspot at home and abroad.The common methods based on vision to target pose measurement,such as four point perspective method,weak perspective camera model method,parasperspective camera model method and direct linear transformation method,these methods need to select at least 3 ~ 4 control points,and the effects are accurate at close range.But when the target is remote,the control points are fuzzy,and difficult to match,that cause the obvious measurement error.Aiming at the above problem,this paper puts forward a target pose measuring method by contour feature matching using Fourier descriptor.The main researches are as follows:(1)The camera imaging mechanism and OpenGL imaging mechanism were analyzed in paper,then this paper proofed the consistency of their imaging mechanism without considering the Open GL frustum shear,and made a theoretical basis for using of OpenGL to obtain the target precise pose image.Then this paper studied the acquirement to 3DS target model,controlling to OpenGL frustum,analog camera observation point and the rotation of the model,and finally get the exact target image at specified model position.(2)This paper expounded the current mainstream methods in contour shape description,and explained the feasibilities in target pose measurement.Then this paper put forward a contour matching method using an improved Fourier descriptor,and this descriptor not only preserve rotation meanwhile eliminate the effect of translation,scaling zooming,but also normalize the contour stating point.Finally,this Fourier descriptor feature in contours matching was verified by simulations experiment.(3)This paper expounded the basic principle of particle filter,then compared with Kalman filter and extended Kalman filter,and indicated the limitations of extended Kalman filter in nonlinear filtering.Next based on the contour feature matching by Fourier descriptor,this paper put forward a pose measuring method using particle filter,then detailed analyzed the state transition,resampling in filter,as well as the processing to starting frame etc.Finally this paper verified the algorithm by digital simulation experiment and physical simulation experiment,then compared the measuring result with feature points matching method.The experimental results show that the algorithm proposed in this paper can accurate measure the remote space target three-dimensional pose,and the accuracy is superior than traditional feature point matching measurement method for remote space target.