| During the process of autonomous relative navigation of two mobility platforms,the position and attitude of the target platform relative to the tracking platform is the key data of feedback for the control system.Vision-based pose measurement methods provide an effective way to obtain those data precisely.In the light of the limitations of existing vision-based pose measurement systems and methods in measuring range,resolution and precision under the condition that the visual distance varies greatly,this paper presents an scheme and implementation of a segmented pose measurement system.The measurement task is divided into two segments according to the scheme.The far-segment measuring module is composed of two groups of orthogonally arranged line-array cameras,which is designed for high-resolution position tracking in the measurement range of 35 m to 5m;The near-segment measuring module is composed of a plane-array camera which is designed for high-precision pose measurement in the measurement range of 0m～5m.In this paper,after realizing system construction including hardware and software design,we focus on the imaging models and measurement methods corresponding to far-segment measuring module and near-segment measuring module respectively.The main research work of this paper is as follows:(1)A segmented pose measurement method based on two groups of orthogonal linear array image sensors and one area array image sensor is proposed,which fully combines the advantages of linear array image sensors and area array image sensors in long-range and short-range pose measurement respectively,and can continuously track the position and attitude of the target with high accuracy in a wide range of visual measurement.A pose measurement system based on this method is designed and constructed,of which the hardware and software are developed.(2)On the basis of the imaging model for line cameras equipped with cylindrical lens,a complete linear 3D coordinate reconstruction method is designed for the far-segment measuring module.In order to eliminate the systematic error caused by lens distortion,a distortion equation of cylindrical lens is fitted by simulation data,which is then used to augment the ideal imaging model of line array cameras.Combining the augmented imaging model with the particular structure of each camera group,we propose a measurement model and calibration method based on orthogonal distortion compensation.This model can solve the problem that the distortion of cylindrical lens can’t be effectively corrected by using the imaging information of one single linear array camera,thus providing an effective solution to eliminate distortion error and improve measurement accuracy of the system.(3)To solve the monocular pose measurement problem for the plane-array camera,two linear pose solutions based on the homography matrix are proposed,which are deduced from algebraic and geometric aspects respectively.They both have analytical expressions,and need extremely low computational cost.Those two linear solutions can be directly used as outputs when the accuracy requirement is not strict,or be used as the initial values for iterative pose optimization.(4)The ambiguity in pose estimation from coplanar feature points is studied from experimental point of view.Based on analysis of the geometric characteristics of the locally optimal pose solutions arose from iterative pose optimization,the assumption "mirror pose" is proposed,which describes pose ambiguity in a concise form with a clear geometric interpretation.Based on this assumption,we improved the pose optimization method so that it can converge to the global optimal solution.Experimental results show that the modified pose estimation method can reach the highest theoretical pose accuracy and stability with relatively lower computational cost. |
PDF Full Text Request