Researches On Fast And High Precision Pose Measurement For Space Objects Based On Feature Points Correspondences

Visual pose measurement method is widely used in aerospace,visual navigation,virtual reality and other fields because of its advantages such as non-contact,moderate accuracy,strong anti-interference and good stability.In order to further improve the accuracy,speed and robustness of vision pose measurement in practical application,this paper studies the pose measurement algorithm based on monocular vision deeply and carries out the work from the following three aspects: first,in terms of the issue that the accuracy of current pose measurement algorithm is limited due to the aperture imaging model error,how to achieve high-precision pose measurement by using the camera general model.Secondly,how to improve the success rate,accuracy and efficiency of pose measurement when the current feature points matching relationship is uncertain.Finally,how to predict the position and attitude of the camera according to video image sequences at the time between frames with high accuracy and high stability,so as to improve the measurement speed in real-time measurement.The main contents of this paper are as follows:1.The theory,expression and calibration method of Incident-ray-tracking model are studied,and a high-precision pose measurement algorithm(PROI algorithm)based on this model is proposed.The proposed algorithm describes the object collinearity error with the more accurate imaging ray in the new model instead of the line of sight in the pinhole camera model,and calculates the object pose relative to the reference plane coordinate system by optimizing the object function iteratively with the orthogonal iteration method.The global convergence of the iterative algorithm is studied,and a method for selecting the optimal initial value of the iterative algorithm is proposed,which reduces the number of iterations of the algorithm.2.The pose measurement with unknown correspondences between image points and object points are studied,and a simultaneous pose estimation and correspondences determination algorithm(Soft OI algorithm)is proposed.The algorithm combines the Orthogonal Iterative algorithm for calculating the pose and the Soft Assign algorithm for determining the correspondence,and establishes the objective function based on the weighted global collinear error.By optimizing the object function iteratively with SVD method and matrix alternate normalization technique,the pose and the correspondences converge to an optimal value simultaneously in a deterministic annealing process.3.The pose prediction of continuously moving targets in space is studied,and a pose prediction algorithm(AUKF Prediction algorithm)based on adaptive unscented Kalman filter is proposed.The algorithm combines camera perspective projection imaging model with Kalman state space model.The image coordinates of feature points are selected as the observed variable,and the pose parameters and the motion parameters are selected as the state variable in the basic unscented Kalman filtering equation.The state equation is used to accurately predict the pose of the target at the sampling interval and the sampling time,and the prior information of the target pose is corrected according to the image coordinates of the feature points.The adaptive adjustment method of filtering noise statistics is studied to improve the accuracy and stability of the prediction algorithm.