Research On Key Technology Of Spacecraft Dynamic Measurement Based On Stereo Vision

During the test of the spacecraft,the explosion of fire artifact,separation of star and arrow,solar wing deployment,antenna deployment,vibration test and impact test,etc.contain dynamic pose testing tasks from low frequency to high frequency.It is very important to select a suitable dynamic pose measurement method for improving the accuracy of spatial pose measurement while achieving the absolute positioning function of the three-dimensional orientation and posture information under the motion state of the measured target space.With the increase of sampling frequencies of camera as well as the processing speed of computer,the stereo vision-based dynamic pose measurement method is widely used in terms of dynamic pose measurements of low-frequency and high-frequency moving targets by virtue of its advantages of wide measurement range,fast measurement speed,high measurement accuracy and real-time measurement,etc.Target dynamic pose measurement is widely used in low-frequency and high-frequency motion.In order to improve the accuracy of dynamic pose measurement of spacecraft under large field of view,this thesis studies the key technology of the dynamic measurement of the spacecraft target dynamic attitude based on stereo vision,which combines with the camera calibration,feature detection and 3D reconstruction in the stereo vision measurement method.Camera calibration,as a critical part of large field-of-view visual measurement,requires large-size targets to ensure the accuracy of camera calibration.Large scale targets are difficult to be produced,and its precision can not to be guaranteed.For a small target,although it has high manufacture precision,only local optimal solutions can be obtained.Therefore,to ensure the calibration accuracy requirement of the large field-of-view,studying the most suitable ratio of the target size to the camera field of view is required.In this thesis,the cameras are calibrated by a series of different dimensions of checkerboard calibration targets and round calibration targets,respectively.These results show that the influence of the ratio of the field-of-view of the planar target on the calibration accuracy,which provide a reference for the reasonable selection of target size for large field-of-view.Aiming at the calibration problem of multi-camera systems,this thesis proposes a multi-camera calibration method based on multi-plane 3D target.Only one captured image of the proposed method can get the intrinsic and extrinsic parameters of multi-camera systems.It solves several problems,including the cumbersome process of using 1D calibration target and the cumulative error while utilizing 3D or 2D target.In the process of feature detection,feature circle and feature corner are difficult to detect in the complicated background.In order to improve the detection stability and accuracy of circle and corner,circle detection algorithm and corner detection algorithm are improved.In the process of 3D reconstruction,the 3D reconstruction method based on the stereo vision mathematical model is improved.Combining the three aspects of works,this thesis writes the dynamic pose precision measurement software based on MFC framework,which has the functions of camera calibration,image processing and data analysis.In order to verify the accuracy and reliability of the software,three experiments are carried out in this thesis.The first and second experiments simulated large field-of-view measurement environment of the spacecraft.The corner points and feature circle are selected as the features respectively,and the high-speed camera is used for shooting.The image features in different positions are reconstructed in 3D,and reconstructed results are compared with the conventional true value.When the measuring range of cameras is 4 m×4 m and the measurement distance of cameras is 4.5 m-5 m,the experimental results show that the absolute error between reconstructed results of corner points and feature circle by software and the conventional true value are within 0.3 mm.In the third experiment,the feature points are posted on the vibrating table,and the vibrating table is set to generate a standard sinusoid with a frequency of 5 Hz and an amplitude of 10 mm.The high-speed cameras are used for shooting,the feature points are reconstructed in 3D by measuring software while the dynamic feature point displacement curve is obtained.The experimental results show that the curve of 3D reconstruction is consistent with the standard sinusoid.