Research On Key Technologies Of Orthogonally Splitting Imaging Vision Pose Measurement System

With the increasing application of vision pose measurement technology in aerospace,military and industrial fields,the spatial measurement requirement focuses on large-range,rapid and high-accuracy.In order to solve the problems of low-speed and low-accuracy of measurement in the monocular vision large-range measurement system with a single area CCD,the principle of orthogonally splitting imaging is studied to achieve monocular pose measurement with high-speed and high-accuracy.The basic theory of orthogonally splitting imaging pose measurement and the key technologies in solving pose measurement have been studied in the dissertion.The main research contents of the dissertion are:1.The principle of orthogonally splitting imaging and pose measurement are studied.The orthogonally splitting imaging sensor acquires a two-dimensional image of the target,which verifies the validity of the orthogonally splitting imaging theory.The nonlinear distortion model of the orthogonally splitting imaging sensor and the distortion correction method based on cross ratio invariance are studied,and the image quality is effectively improved by the distortion correction.2.A calibration method of global imaging model based on global radial basis function interpolation is proposed,which overcomes the influence of nonlinear distortion on the measurement accuracy and improves the calibration accuracy.This method is suitable for complex application.The mathematical model of the orthogonally splitting imaging sensor with Plucker line and radial basis interpolation method based on the general imaging model is established.An adaptive selection method based on K-means clustering is proposed to select control points adaptively in radial basis interpolation method.According to the characteristics of Plucker line,the calibration formula of the camera matrix is derived and the matrix parameters are solved by the least squares method.The effects of different radial basis functions and different shape parameter values on the calibration accuracy are studied.3.Based on the general imaging modelan improved calibration method-Partition of Unity Method based on KDFcm(Fuzzy C-Means Clustering Based on Kernel Distance)is proposed.The method can solve the problem of ill-conditioned dense matrix in the calibration method based on the global radial basis function interpolation method in dealing with large-scale data.The Partition of Unity Method based on the KDFcm is studied.The method of constructing the weight function of the orthogonally splitting imaging sensor with a locally compactly supported radial basis function is proposed by using the Shepard method.The influence of the Partition of Unity Method with different radius expansion coefficients on the calibration accuracy is studied.4.In order to solve the problem that so many geometric constraints affect the pose accuracy in the process of pose solution,a non-linear unconstrained optimization measurement model of the orthogonally splitting imaging vision pose measurement system is proposed.The penalty function is used to combine multiple geometric constraints,and the constrained optimization problem is transformed into an unconstrained optimization problem,and a nonlinear unconstrained optimization measurement model is constructed.The camera coordinates of feature points are solved iteratively.The displacement and angle calculation algorithms are studied.The measurement accuracy of different calibration methods for pose measurement is compared by experiments.The displacement measurement accuracy is ±0.1mm,and the azimuth angle measurement accuracy is ±0.1°.And the measurement accuracy of the algorithm in this paper is higher than the non-linear geometric constraint measurement algorithm through experiments.