Research On Position And Pose Estimation Of Flight Target In Range Based On Binocular Vision

The three-dimensional pose of flight targets in range is an important technical indicator reflecting their flight status.The precision measurement has very important practical application value in the aspects of test analysis,accident identification,training evaluation,test of flight control system and design of new weapons.Vision-based pose measurement method has the advantages of non-contact,large measuring range and high accuracy.The photoelectric theodolite is used to track the flying target in range and take sequence images of the target.The image interpretation methods are applied to the image sequences to get the high-precision measurement results of the target’s three-dimensional position and pose.There are still some problems in current image interpretation methods for position and pose measurement of flight targets in range.Firstly,the difficulty of feature extraction during the image interpretation process increases when position and pose measurement is performed on moving targets without the cooperation marker.Secondly,because of large change of attitude in the flying process and the serious self-occlusion,it is difficult to track targets’ features accurately and robustly.Finally,the distance between the theodolite and the target is relatively large,so the target area in the acquired image is small,which reduces the accuracy of feature extraction.In view of the above problems,this paper studies the method of pose measurement based on binocular vision.The main work of this paper is as follows:Based on the central perspective imaging model,use OpenGL and 3D Studio Max(3DS Max)software to simulate the photoelectric theodolite acquisition target image sequence process,the simulation of the flight target sequence images generated for the test range flight target pose measurement algorithm.A pose measurement algorithm based on parallel line clustering is proposed for straight-wing aircraft.The algorithm does not need other priori information,such as 3D model,cooperative mark points and so on,it only needs a pair of images taken at the same time to estimate the position and pose information.In the algorithm,the main structure of the straight wing aircraft is extracted through line detection and parallel line clustering,and the position and pose information of the flight target is calculated using line-line intersection and plane-plane intersection.A pose measurement algorithm based on vector analysis is proposed for swept-wing aircraft.The projectile invariance of the wing-body is proved against the swept-wing aircraft under the perspective of the weak perspective projection model.Firstly,the algorithm extracts the fuselage by using the parallel line clustering algorithm,and then uses the invariance of the projection relation to extract the two front edge lines of the wing.Finally,the position and pose information of the flight target is calculated using line-line intersection and plane-plane intersection.For flight target whose 3D model is known,a pose measurement algorithm based on 3D projection is proposed,which is suitable for various flight targets.Firstly,the algorithm selects several points on the three-dimensional model,at the same time uses the optical flow method to track the feature points in the image to determine the initial value of the target position and assuming the initial value of the target attitude.Then project the selected 3D points of the model to the 2D image,detect two convex hull for tracked feature points and 3D points projection.Finally,the result of the pose estimation of the flight target is obtained by the convex hull align criterion.Based on binocular vision pose measurement principle,a pose tracking algorithm of target tracking based on feature tracking is also proposed.The feasibility and accuracy of proposed algorithms are verified by experiments in simulated images and real images.