Star Image Recognition And Its Application In Attitude Determination For Space Crafts

Attitude determination is a significant project to guarantee the normal operation and mission success for space crafts.Currently star sensor is a type of autonomous attitude determination equipment with the highest accuracy.In order to realize navigation of high performance with INS/CNS integrated navigation system which consists of inertial measurement units and star sensors,it's necessary that multiple procedures such as star recognition,attitude calculation,information fusion and so on are synthesized.The paper researches algorithms for full-sky autonomous star recognition,fast algorithms for attitude calculation and the effect of star sensors in INS/CNS integrated navigation system in the course of autonomous attitude determination of star sensors.As follows are the contents included:First,the algorithms for full-sky autonomous star recognition without prior attitude are researched and two improved algorithms for full-sky autonomous star recognition are proposed.The algorithms are named as star recognition algorithm based on polygon division and star recognition algorithm based on path optimization with randomly distributed ant colony.The former algorithm uses a kind of polygon which certifies their own sizes according to the distribution of stars in the star images.The star image is divided into a set of polygons and the feature pattern of the star image is extracted through geometrical features of the polygons.Compared with the typical grid algorithm,the position information of the stars in the image is used more adequately.It's not necessary to determine the size of the image units and the feature pattern is invariant to rotation.When the position noise probability increases to 0.5,the algorithm can still presents a recognition ratio higher than 90%.The latter algorithm defines the ant colony which is randomly distributed through simulating the biological characteristics of ant colony in nature.The algorithm certifies the optimal traversal path through all the stars in the star image,rather than choose the star nearest to the image center as the starting point for path optimization.Thus the possible error caused by starting point determined in advance is eliminated.The simulations indicate that this algorithm presents a successful recognition ratio which decreases less than 7% when the position noise probability reaches 0.7.It presents stronger stability under the image noise.Second,the algorithms for calculating the instantaneous attitude in star sensors are researched.An algorithm for calculating the instantaneous attitude based on reversible linear transformation is proposed.The attitude equation set which takes the attitude quaternion as unknowns presents its particularity.According to this particularity,the equation set is transformed into standard quadratic form through reversible linear transformation.It can reduce the calculation quantity to solve the equation set after transformation,which is especially suitable for the situation of the guiding star shortage.The comparisons with several attitude calculation algorithms indicate that,the calculation time decreases by 40% through the proposed algorithm on the premise to ensure the calculation accuracy.The real-time performance for attitude output is enhanced.Third,the information fusion methods in INS/CNS integrated navigation system are researched.Aimed at fusing the space craft's attitude output from INS and the attitude information measured by the star sensor,an improved nonlinear filtering algorithm named as nesting particle filtering algorithm is proposed.The basic particle filtering algorithm is utilized to generate the mean value in the density of importance sample for the predicting particles and the unscented Kalman filtering algorithm is utilized to generate the variance.Thus the new predicting particles can be generated.The particles are returned to the step of generating particles in particle filtering algorithm and the filtering starts.The practical calculating example indicates that,the improved nonlinear filtering algorithm can provide higher filtering accuracy than particle filtering algorithm and can overcome the disadvantage of determining the density of importance sample.Meanwhile,the simulations indicate that the improved algorithm for star recognition enhances the navigation accuracy in integrated navigation system applying the filtering algorithm above.Fourth,the integrated modes in INS/CNS integrated navigation system are researched and a switchable deep integrated mode is proposed.If the observed stars in the star sensor's field of view are not enough to guarantee the accuracy of the information provided by CNS,the integrated navigation system will be switched to the deep integrated mode.After the measured variables of the integrated system are estimated by the filter,the attitude angles from INS are transformed and utilized to correct the attitudes calculated from the star sensor.Meanwhile,the measured information output by CNS keeps correcting the navigation parameters of INS.Simulations are performed to validate this integrated mode.The results indicate that through the deep integrated mode,navigation with high accuracy can be realized.Meanwhile,the simulations indicate that the improvement can also enhance the performance of navigation.