| Satellite attitude measurement and determination system is one of the key technologies of satellite,and the combination of star sensors and gyroscopes is a common way of attitude determination in satellites.With the continuous improvement of satellite attitude accuracy,it is urgent to strengthen the research on high-precision star sensor and integrated attitude algorithm.Aiming at the key theory and technology on satellite attitude determination based on star sensor and gyro,this paper make deep research on relative technologies such as star extraction,star pattern recognition,on-orbit calibration model and integrated attitude algorithm,and test them with simulation and experimental experiments.The main research contents and achievements are as follows:1.Star extraction.Large amount of work is imposed on the threshold segmentation algorithm.Based on the traditional global threshold segmentation algorithm,a new improved threshold segmentation algorithm is proposed in this paper.The algorithm takes the average background of multiple star maps as the threshold template,and adjust the threshold template according to the statistical characteristics of the threshold template and the map to be handled,which achieved a good separation of background and target.A new star extraction process is designed and realized as well.Using the traditional threshold algorithm could only extract about 20 stars(for the 20.3°FOV,6.5 mag sensitive threshold star sensor),and by using the improved threshold algorithm and the new star extraction process,not only can we remove all kinds of image noise,but also improve the number of extracted stars to about 100 averagely,which largely improve the amount of data for star map identification,on-orbit calibration and attitude determination.2.Star pattern recognition.Two star pattern recognition algorithms are studied in this paper – the star pattern recognition algorithm based on quadrangle and star pattern recognition algorithm based on K vectors.Firstly,the star map recognition algorithm based on quadrilateral is studied.The advantage of this algorithm is that the redundancy is small,the disadvantage is that the computation is large and the matching time is long.Then,based on the principle of K vector search and identifier recognition,a new fast star pattern recognition algorithm is proposed.The algorithm uses K vector search method to quickly match the angular distance,and the correspondence relation between stars is identified according to the identifier recognition principle,which get four basic stars.Then complete the whole field of view match based on the four basic stars.The star pattern recognition algorithm in tracking mode is also designed in this paper.Lastly,the algorithm is applied to the star pattern recognition of real ZY-3 satellite,and results show that the algorithm has characteristic for quick working speed and high success rate,which can meet the need for star sensor.3.On-orbit calibration model for star sensors.Focused on the on-orbit calibration for large field view star sensor,a new improved on-orbit calibration model basing on vector observation is proposed in this paper.The model used three Rodriguez parameters instead of three Euler angles as the exterior orientation parameters,and completed the joint solution with the inner orientation parameters and third order distortion parameters.The simulation shows that the model has the advantages of high precision,simple parameter setting,no prior attitude information and good stability in compared to the models with Euler angles.In the case of a standard deviation of 0.05 pixel star point error,the model can still accurately calculate the camera parameters,the statistical deviation of the star angle after the calibration is 0.228 arcsecond,and the RMS value of image projection residual were control in 0.0282 pixel in x direction and 0.0275 pixel in y direction.In the case of large dip angle,the traditional model with Euler angle was not stable and could not converge,but the new model could steadily converge to the vicinity of the preset parameters.The calibration model are also applied into the calibration of ZY-3 satellite star sensor and a land-based astronomical navigation system(with non-professional lens of about 52°×52° FOV),the results show that the statistical deviation of the star angle?Root-Mean-Square value of image projection residual in x direction and y direction are 1.868 arcsecond?0.142 pixel?0.157 pixel and 3.574 arcsecond?0.171 pixel?0.1206 pixel respectively.4.Integrated attitude algorithm.The two main aspects are studied in this paper: Firstly,the traditional extended Kalman filter(Extented Kalman Filter,EKF)and unscented Kalman filter(Unscented Kalman Filter,UKF)are applied in the Satellite Attitude Determination System based on star sensors / Gyros,and the Cubature Kalman filter(CKF)raised in recent years is also analyzed and implied.Compared the performance of these three algorithms and made precision analysis on integrated attitude determination system.Secondly,considering the fact that the integrated attitude determination system may encounter some problems such as uncertain system model,breaking state and bad measurements in the space,a novel satellite attitude determination algorithm--Strong Tracking Adaptive Square Root Cubature Kalman filter(STSRCKF)is proposed in this paper.To get a better adaptability as STF in handling with uncertain system model and breaking state problem,a fading factor is introduced into the algorithm.To ensure the positive definiteness and symmetry of the covariance matrix,a square-root decomposition method is adopted into the calculation of the covariance matrix,which improved the filter's accuracy and stability.To get a better adaptability in handling with bad measurements,anomaly detection process and a self-adapting factor is introduced into the algorithm.Simulation experiments shows that STSRCKF has higher accuracy,better robustness,stronger stability and tracking performance than any single filter in dealing with uncertain system model,breaking state and bad measurements.5.The measured data processing.Two experiments are carried out in this paper.The first experiment is carried out by using a land-based astronomical navigation system,and a set of processes and methods,such as experimental data acquisition and processing,precision evaluation,are designed.Experiment show that static attitude acquisition accuracy of the system is better than 4 arcsecond(1?).Second,based on 800 star maps of ZY-3 satellite,the space-based dynamic attitude experiment is carried out.The star extraction algorithms,the star pattern recognition algorithms,the on-orbit calibration models for star sensor and the definite attitude determination algorithms are further verified,which illustrates the practical value of this paper.However,due to the lack of matching data of star maps and gyros,we can only verify the combination of attitude determination algorithm from the perspective of simulation,but also provided some reference value for the actual engineering application and data afterward handling of remote sensing satellite. |
PDF Full Text Request