Research On Attitude Determination Methods Of Spacecraft Based On Compressive Sensing

The star tracker is the most accutate sensor which can determine the absolute three-axis attitude of a spacecraft.As the star tracker develops towards wide field-of-view,high accuracy and high dynamic directions,high-accuracy star tracker suffers the problem of high data amount and insufficient processing rate,which largely restricts its efficiency and application.The emergence of compressive sensing(CS)theory provides a novel approach for solving these problems by changing the traditional imaging method.This dissertation mainly focuses its research on CS application in attitude determination of star tracker containing three key factors such as sparse representation,incoherent measurement and sparse reconstruction,and investigates the reliable star identification and attitude determination methods with respect to compressive imaging.The main results achieved in this dissertation are summarized as follows:First,the sparse representation performance of star image is analyzed quantitatively.The basic sparse representation theory is introduced,which contains the common sparse representation approaches by complete basis,overcomplete dictionary and learned dictionary.According to the evaluation index of sparse representation for star image regarding the image level and the feature level,the sparse representation performance in different sparsity and representation bases is quantitatively investigated with respect to star image error and star centroid error.The results prove that the sparsity of star image meets the requirement of compressive sampling,and verify the feasibility for applying compressive sensing in attitude determination by star tracker.Second,a weight block circulant matrix(WBCM)that is suitable for compressive measurement of star image is formed,and a compressive measuring method of star tracker based on deterministic phase modulation(DPM)is presented.The WBCM can enhance the sampling in low-frequency part,and obtains a high performance on compressive measurement to the signal that contains comparatively large low-frequency information such as star image.Meanwhile,according to the imaging requirement of star tracker,the WBCM is implemented in a DPM based compressive measurement method.The results verify the feasibility of DPM method,and show that this method has better performance on compressive measurement than the compared traditional methods.Also,the DPM method is proved to be easy to implement into hardware and represents an attractive perspective in application.Third,a compressive measurement model based on spatial sparsity of star image is constructed,and a feature-focused reconstruction method of star image data is proposed.The effect of sparse reconstruction on star image is analyzed theoretically.According to the evaluation index of reconstruction error of star feature,the reconstruction errors of the star centroid,brightness and mistaken stars are analyzed quantitatively.The results show that the reconstructed star can keep its features for attitude determination to a large extent.Besides,a feature reconstruction method of star image based on optical superposition is proposed,which is based on the spatial sparsity of star image and the measurement model of compressive superposition.The advantage of this method is that it extracts the star centroid feature from compressive data and directly reconstructs the feature,which represents an innovative feature-focused reconstruction method.The simulations show that this method can reduce the amount of data processing and reconstruction time effectively,and the results prove its validity in solving process and robustness in noise cases.Fourth,the error models of singular value decomposition(SVD)algorithm are formulated,and an improved method of high robustness is proposed.The inherent relationship between singular value and star vector in SVD algorithm is studied,and the boundaries of three singular values are derived in general cases.Considering the reconstruction error of star image,the the error models of SVD algorithm in terms of position error,magnitude error and mistaken stars are formulated respectively.Based on above conclusions,an improved SVD method of high robustness is proposed,which overcomes the influence of reconstruction error on star identification and attitude determination to a large extent.The simulations verify the correctness of the derived singular values' boundaries of pattern matrix,prove the validity of the improved strategy,and show that the improved SVD method obtains better robustness against the above errors than traditional method.The results guarantee the accuracy and the feasibility of star identification and attitude determination through reconstructed data.Finally,the attitude determination by compressive sensing is verified by comprehensive simulations.The results show that the approaches based on compressive sampling and reconstructed data processing can implement the attitude computation in the correct way,and keep the accuracy of attitude determination to a large extent.