Research Of The Feature Points Location Of Cooperative Target Based On Hand-eye Vison

Robotic arm is an important component part of the International Space Station,which can assist astronauts to accomplish many operating tasks in outer space thatneed high precision, such as recognition, tracking and grabbing of space target. Thetarget images are taken by the hand-eye camera and the position and attitude of thespace cooperative target can be measured precisely by the hand-eye vision system ofrobotic arm. With the development of the space technology, the accuracy requirementof the location of space target is getting increasingly higher. Since the measurementaccuracy of the position and attitude of the target is directly affected by the locationaccuracy of the feature points on the cooperative target, it will be of great significanceto research algorithm of the feature points location on cooperative target by hand-eyevision system.The algorithm of the location of feature points on cooperative target is roughlydivided into three stages. The first stage is the recognition of the cooperative targetarea from the target image. Then the feature points used to calculate the position andattitude of the target will be recognized from the target image area. At last, it is thelocation of the feature points on target. The last two stages were deeply researched inthis paper.This paper introduces the principle of the measurement of the position andattitude for cooperation target based on hand-eye vision. Through experiments it isverified that the location accuracy of the feature points on cooperative target isimportant for the measurement accuracy of the position and attitude for the target.Then the existing location algorithms of feature points are analyzed and comparedwith each other through experiments. The analysis and the experiments show that theedge detection operators usually used in the traditional center location algorithms for feature points are sensitive to noise, which leads to that the edge detection is neitheraccurate nor continuous, and it will have a negative effect on the subsequent edgefitting precision. To solve this problem, the paper proposed a location algorithm forthe target center based on geometric features. Firstly, coarse location of circle targetcenter is realized by combining adaptive threshold segmentation and centroidmethod, which is used for radius constraint of the edge detected by Canny operator toremove the isolated points and the noisy points. Then, starting from the geometricdistribution characteristics and linking discipline of the ideal imaging of circle，amethod based on partition theory is brought forward to obtain ideal and continuousedge of circle target. Finally, Zernike moment is used to carry out the subpixellocation of the edge, and the center location is achieved by the least-squares ellipsefitting method. Results of experiments show that the location acc uracy(3δ) can reachto0.0711pixels，and that the operation time is between2and3ms. It can satisfy thehigh requirements of precision, stabilization and real-time ability of circle targetcenter in the measurement system.Because of the complexity of the space lighting condition, image processing canbe easily affected by the two factors, uneven lighting and incomplete target. Then thecoarse location of circle target center should not adopt the adaptive thresholdsegmentation method. A novel algorithm of feature points recognition was proposedin this paper using the geometry features of target and position relation of threefeature points. Iterative least squares fitting method was used in the stage of coarselocation of circle target center, removing the illegal points. Continuous edge of circletarget was obtained by the edge-tracking method based on partition theory. Theexperiment results show that the location of the feature points on cooperative targetcan be high-precision in both cases as mentioned.