Research On Autonomous Recognition And Pose Estimation Of Non-cooperative Spacecraft Based On Point Cloud

With the rapid development of space technology,more and more spacecrafts are launched and space on-orbit service has become a research focus all over the world.On-orbit service includes spacecraft life extension and maintenance,fault satellite repair and recovery,space debris removal and deorbit,which is the key to ensure the security of space mission and maintain the space environment.At present,most of the on-orbit spacecraft are non-cooperative targets so autonomous recognition and pose estimation for non-cooperative spacecraft is the premise and foundation of space on-orbit service.Firstly,the advantages and disadvantages of vision measurement and laser measurement are compared and laser radar is used to acquire point cloud data of non-cooperative spacecraft.In view of the acquired point cloud,three point cloud reduction methods including uniform grid method,curvature sampling method and grid barycenter method are compared.The grid barycenter method is used to compress the point cloud data and reduce the memory pressure of onboard computer.Comparing radiusbased filtering method with statistical-based filtering method,the statistical-based filtering method is used to eliminate the noise in the point cloud and provide high quality data for the follow up operation.Aiming at the characteristic that non-cooperative spacecraft have no cooperative marker and grasping mechanism,an autonomous recognition method without prior information for non-cooperative spacecraft is proposed.Firstly,the normal information of point cloud is introduced and mainstream normal estimation methods are compared.The weighted covariance analysis method based on plane fitting is used to estimate normal and structure point cloud topological relation,in order to prepare for subsequent feature recognition,point cloud registration and pose estimation.Then the basic principle of the traditional RANSAC algorithm is introduced and RANSAC algorithm combining normal information is adopt as the basic scheme of recognition.Based on analysis of geometric characteristic of non-cooperative spacecraft,specific recognition methods are designed for typical spacecraft components such as solar panel,spacecraft body,signal transmitting antenna and engine nozzle.The experimental results show that the proposed method is applicable for recognition scenarios without any prior information and can satisfy the requirement of autonomous recognition task for various types of non-cooperative spacecraft and is the critical step with robustness and universality of pose estimation and target capture.In view of the characteristic that unknown motion state of non-cooperative spacecraft and lack of information exchange with serving spacecraft,a pose estimation method for non-cooperative spacecraft without prior information is proposed.Firstly,the measuring coordinate system is established and definition of pose transfer matrix is given to prepare for the subsequent position and attitude estimation.Based on the result of autonomous recognition of spacecraft,the identified components are used as base model and point cloud at different time are registered by ICP algorithm.The transfer model of pose matrix is established and continuous motion tracking and pose estimation of non-cooperative spacecraft are completed by correlating the transfer parameters of adjacent point cloud frames.The experimental results show that the proposed method doesn't need the shape and size of the spacecraft which is faster and more accurate than traditional whole spacecraft pose estimation method.It can meet the requirement of accuracy and real-time of on-orbit service operation for non-cooperative spacecraft.In the end,in order to verify the effectiveness of proposed method,an experimental platform for autonomous recognition and pose estimation of non-cooperative spacecraft is designed.The Kinect V1 sensor is used to acquire the point cloud of spacecraft,the satellite model is installed on electric translation stage and electric rotating platform to simulate motion of spacecraft,and personal PC is used to simulate the onboard computer for follow-up data processing.In addition,the software for autonomous recognition and pose estimation of non-cooperative spacecraft is developed,which makes user actions more convenient and provides technical support for space on-orbit service.