Research On The Technology Of Star Image Preprocessing And Motion Blur Restoration

Nowadays with the rapid development of micro and small satellites,star sensor is the best choice to measure the attitude of satellites in view of that gyro is too unwieldy while the star sensor is high accuracy,light weight,low power consumption and small size.The goal of this dissertation is to advance the precision of the star sensor with the help of star image preprocessing technique and image deblurring.In this dissertation,the satellite orbital dynamics model,attitude kinematics model and the related coordinate systems,descriptions of the attitude are introduced.The fourth-order Runge-Kutta method is wielded to obtain the satellite orbit model,at the same time,the three-dimensional orbit is simulated using MATLAB.The conversion between Euler angles and quaternions in the attitude description methods is verified.The principle of projection transformation from the reference coordinate system to the star sensor imaging coordinate system is dissected and the simulated star image is generated according to the star energy distribution feature and the magnitude and intensity of illumination conversion relationship,which meets the requirements of experiment and research.And the simulation software of star sensor attitude determination system is designed and implemented.Considering the source and characteristics of the star noise,this thesis puts forward a new approach named Modified Strip Filtering algorithm(MSF)to suppress the strip noise in the star image,whose cumulative centroid error is 21.18% lower than the HVSF algorithm.Moreover,a dynamic threshold denoising algorithm is used to decrease the background noise.Furthermore,for the single point and small area noise,a median filtering algorithm based on conditional judgment and statistical information is proposed,which avoids the loss of the star information while filtering the noise.After removing the noise,aiming at the problem that bad elements may exist in the range of star pixels,this work comes up with a mean filtering algorithm based on sparse expression and threshold to detect and compensate it,which makes the accuracy of the star centroids and the efficiency of data processing improved dramatically.The mechanism of motion blur in the star image is studied,and a simple model is applied to estimate the fuzzy parameters.Meanwhile,several commonly used motion blur recovery methods are analyzed and compared,and the pros and cons of restoration star images are evaluated by taking advantage of image quality performances and star centroid accuracy.Aiming at the issue of noise accumulation in Lucy-Richardson algorithm,this thesis brings forward an improved algorithm based on template and dynamic windows,which improves the anti-noise performance of the algorithm and reduces the influence of noise on the restored image quality.