Research On Mono-vision Pose Measurement For Space Cooperative Target

Space optical loads’ aerospace activities are significant research directions in recent years,such as on-orbit manufacturing,assembly,maintenance and repair,space transportation and rescue,space debris or failure satellite cleaning,spacecraft synergies and cooperation.The method of visual measurement is usually used in the near range approximation of the spacecraft for rendezvous and docking.It has become a hot spot of research since the method has many advantages,such as simple structure,stable performance,low cost and so on.This paper mainly studies the mono-vision pose measurement technology of spatial cooperative target.Its core contains a series of image processing operations of cooperative target image,obtained by monocular camera,such as recognition,feature detection and positioning.PnP algorithm is used to solve the relative pose between camera and target.The design of the cooperative target is the first step in the mono-vision measurement.The overall design of the plane four point and code information is determined through the comparison and analysis of the existing schemes.Considering of the working parameters of the camera,the relationship between the distance of feature points and the accuracy of pose is derived,and the parameters of the detailed size and the characteristic arrangement of the target are determined.Finally,a cooperative target with a wide range of functions,high precision and strong robustness is designed.The recognition of the cooperative target is the second step in the mono-vision measurement.Continuous,smooth,single-pixel-wide edge is obtained by using the principle of pixel gradient and chain code in preprocessed image.The outer frame,the circle and the line segment are identified by the constraints of the roundness,the inflection point detection and the center matching.Then,target is identified by the decision condition.The algorithm can identify the cooperative target steadily,accurately and quickly in complex scenes such as different distance,diversity of attitude,local occlusion,complex illumination and so on.The location of the feature points is the third step in the mono-vision measurement.The location algorithm of frame intersection and circular center based on least-squaremethod and statistical-regression-method are designed at the far and near location respectively,which ensures the accurate,fast,sub-pixel positioning of feature points.By using the affine invariance of the ratio between the area of the inner circle and the area of the outer circle,the coded information of the target can be decoded successfully.The algorithm can be adapted to all kinds of pose and the result is unique,and so as to prepare for the pose calculation in the next step.Pose solution is the fourth step in the mono-vision measurement.In this paper,the theoretical derivation and algorithm implementation of the P4 P algorithm are carried out to solve the relative pose of the camera and the target.The final step is to verify the accuracy of the visual measurement.Because of the high accuracy of visual measurement,it is difficult to calibrate the relative pose in high precision between the camera and the target by using other instruments.Therefore,the UR10.0 robot is used to build a precision analysis system to analyze the error of the measurement results.Using the upper computer,the remote monitoring and control of the UR10.0 robot is realized through the TCP/IP protocol and the URScript language.Finish the hand eye calibration of the robot and the target.After finishing eye-in-hand calibration of UR10.0 robot,experiments are designed to verify the repeatability accuracy and absolute accuracy of the visual measurement.The results meet the requirements of the subject.As a whole,the processing speed of the visual measurement system is about 84.3ms/frame,and the accuracy rate of target recognition is 95%.The measurement error is less than 0.05 mm and 0.05° at the nearest distance 0.13 m,and is less than 5mm and 1° at the farthest distance 2.2m,which meets the requirements of speed and accuracy.At present,the research results in this paper have been successfully applied to operations of docking,locking and releasing to simulate the tasks of space rendezvous and docking.It lays the necessary technical foundation for on-orbit assembling space optical facilities in the future.