Star Identification Method For Three FOVs Star Sensor

The star sensor is an important part of autonomous navigation system of spacecraft due to its autonomy and high accuracy of attitude determination.However,because of the limitation of the field of view(FOV),the single FOV star sensor will be invalid when it captures less than three stars,which affects its continuous navigation.The problem can be solved effectively by multiple FOVs star sensor,of which the star identification method is a core technology.The star identification method for three FOVs star sensor is studied in this thesis,and the main work and contributions are summarized as follows:Considering the conventional triangle algorithm has poor performance in robustness,time consumption and storage space,a new generalized regression neural network(GRNN)based star identification algorithm for single FOV star sensor is proposed.Firstly,the double stars,variable stars and the star magnitude in the SAO J2000 fundamental star catalogue are processed.Consequently,the navigation star catalogue is built.Secondly,every navigation star in the navigation star catalogue is characterized with a triangular pattern,which is used as the input of GRNN,and then the star identifier is trained.Finally,observed star images are simulated based on the navigation star catalogue and identified by the star identifier.The simulation experiments show that the GRNN based star identification algorithm holds high accuracy of star identification,high efficiency and strong robustness.Aiming at the issues that the single FOV star sensor's roll angle accuracy is not as precise as other two angles and the reliability of star imaging is apt to be influenced by environment,three FOVs star sensor is studied and the two-step high accuracy star identification algorithm for three FOVs star images is presented.At the first step,the GRNN based star identification algorithm is utilized to identify the star images that capture more than three stars and the identification results are checked.If the number of correctly identified images equals to one or two,the rough attitude of spacecraft will be calculated,otherwise,if three star images are all successfully identified,star identification will be completed and the precise attitude of spacecraft will be determined.At the second step,the bore-sight directions of unidentified FOVs are estimated by the rough attitude of spacecraft,which assist these star images to accomplish triangle algorithm star identification.Based on the two steps above,precise attitude of spacecraft is determined by the star vectors from three FOVs.In order to validate the algorithm,simulated three FOVs star images are made by combining the navigation star catalogue with the installation of three FOVs.The simulation results illustrate that the two-step high accuracy star identification algorithm for three FOVs star images has high identification rate and speed,moreover,three FOVs star sensors have higher attitude accuracy than single FOV star sensors.The satellite tool kit(STK)is used to generate the flight state data,and Matlab GUI is utilized to build the star identification simulation test system for the Lunar Orbiter Spacecraft during its earth-circular phase and the earth-moon transfer phase.This system mainly consists of star image simulation module,star identifier training module,star identification module,attitude determination module,parameter setting module and performance analysis module.The star images are simulated on the basis of 938 samples of attitude data in earth-circular phase and earth-moon transfer phase,which demonstrate and validate the algorithms proposed in this thesis.