Research On Non-cooperative Target Pose Measurement Technology Based On Binocular Vision

In recent years,the aerospace field has made rapid progress in rendezvous and onorbit services.In space docking missions,cooperative targets are generally set on the surface of the spacecraft to assist the completion of the pose measurement.However,in the tasks of cleaning up space garbage,recovering failed satellites and refueling the space station,the unknownness of non-cooperative targets puts forward higher requirements for pose measurement tasks.Computer vision technology is the key to realizing non-cooperative target measurement tasks.technology.Based on the principle of binocular vision,this paper takes the rectangular solar panel of the spacecraft as the research object,designs a set of measurement system with strong stability and high precision,and verifies the effectiveness of the measurement system through two sets of simulation experiments.In the measurement system,indepth research is mainly carried out from the following aspects.(1)Based on the engineering requirements of ultra short range navigation,aiming at the non-cooperative target measurement technology and binocular vision measurement technology,the research status and research significance at home and abroad are systematically investigated and analyzed,and the overall research ideas and technical routes are determined.(2)The camera calibration method is studied,and the classic Zhang Zhengyou calibration method is introduced.On this basis,a binocular vision system is built,and the camera is calibrated by an international general software(M software),and then the internal parameter matrix,external parameter matrix and distortion coefficient of the camera are obtained.The calibration experiment controls the average reprojection error.at 0.11 pixels.(3)In image preprocessing,the denoising effects of spatial filtering and wavelet transform are compared,and a preprocessing method combining wavelet transform and image enhancement is designed.In terms of feature extraction,the traditional FAST algorithm has been improved to reduce the influence of illumination on the FAST feature point extraction effect under different attitudes,and the point feature and line feature are combined to make up for the difficulty of extracting in a dark room environment to a certain extent.Defects of image features,and retain the advantage of high accuracy of measurement algorithms under point features.(4)In terms of stereo matching,the normalized cross-correlation coefficient(NCC)matching method is combined with epipolar constraints to narrow the search range of matching points from the two-dimensional plane to the epipolar line.The efficiency of matching is greatly improved,and the accuracy of matching algorithm is also improved to a certain extent.In addition,a post-processing method of feature matching is also proposed to reduce the probability of mis-matching of the same point.The accuracy and stability of the measurement algorithm are verified by two sets of simulation experiments under normal visible light conditions and dark room conditions.(5)In order to improve the real-time performance of the algorithm and reflect the integrity of the measurement system,the measurement algorithm is transplanted to the Hi Silicon Hi3559 hardware platform.Image processing is completed through the VI module,VPSS module and VO module in the Hi Silicon platform,and the IVE module is used to accelerate the algorithm.