Non-cooperative Spacecraft Structure Recognition And Pose Measurement Based On Binocular Vision

As the pace of space exploration accelerates,the number of satellites launched each year worldwide continues to increase.Spacecraft on-orbit service technology has become an important means for extending the life of spacecraft,maintaining equipment,and removing space debris.At present,most of the orbiting spacecraft are non-cooperative targets,ie pose information cannot be exchanged and no identifier for auxiliary measurement is installed.How to accurately identify the main non-cooperative structures and obtain accurate pose information is the key to achieving non-cooperative targets on-orbit service technology.Therefore,this dissertation mainly studies the problems related to the main structural recognition and relative pose measurement of the non-cooperative target in the binocular vision system.The binocular vision system designed for non-cooperative target structure recognition and relative pose measurement can be mainly divided into the following six parts: image segmentation,structure recognition,edge processing,feature point extraction and matching,three-dimensional reconstruction,and size bits.Posture information solution.In this paper,in order to achieve the target and interference background segmentation,an accurate background model needs to be established.In this paper,the method of removing adjacent frame targets is designed to build the current frame image background model;based on the feature differences between the non-cooperative target structures,the design is based on K.-Means clustering structure segmentation method and structure recognition method based on geometric features;In order to obtain smooth and complete edge images of each structure,this paper adopts morphological method to process the structure image;in order to obtain accurate pose information,this paper aims at the binocular vision.The system performs accurate mathematical modeling,and designs a corner extraction algorithm based on Hough line detection for rectangular targets.An attitude calculation method based on the rotation principle is designed for elliptical targets.Finally,a binocular vision ground simulation system was built,including a movement simulation system and a binocular vision measurement system.Matlab was used to program and the binocular vision system designed in this paper was simulated.The process of noncooperative target structure recognition and measurement based on binocular vision system was simulated through experimental simulation,and the identification and measurement results of the main satellite structure were tested.The experimental results show that the designed system has the advantages of high recognition speed and high measurement accuracy.