| Inertial Navigation System(INS)and Celestial Navigation System(CNS)integrated navigation system is widely used in high-altitude aircraft,such as high-altitude sounding rocket,high-altitude reconnaissance UAV and so on.However,there is still a lack of research on navigation star image processing,navigation star identification,and navigation parameters estimation of the CNS.Based on the research of data processing and measurement acquisition of the CNS,a multi-sensor aided high-accuracy celestial navigation parameters estimation strategy is proposed in this paper.The accurate CNS navigation parameters is used to correct the INS state error and improve the INS/CNS integrated navigation performance.The main research contents are as follows:Due to the influence of device attributes,cosmic rays,and working environment,the navigation star images will be polluted by the Gaussian-Poisson mixed noise.However,each traditional denoising method only has good denoising effect for specific noise.For this issue,a sparse representation denoising method for the Gaussian-Poisson star image noise is proposed.Based on a large number of star image samples,a sparse dictionary is first trained.Then,the captured star image’s sparse coefficient is estimated in real time.The denoised navigation star image is reconstructed by using the sparse dictionary and the sparse coefficient.The simulation results show that when the Gaussian-Poisson noise parameters are 0.08 and1/2 respectively(upper bound of the simulation),the reconstructed navigation star imgae’s peak signal-to-noise ratio reaches 30.21 d B.Missing detection of high-magnitude star points in the denoised navigation star image will lead to the failure of the traditional navigation star identification methods,which only use the stellar angular distance as the identification feature.Aiming at this issue,a navigation star identification method based on stellar magnitude is proposed.The proposed method constructs a Gaussian distribution for each navigation star in the Smithsonian Astrophysical Observation star catalog(SAO),whose energy distribution is determined by the corresponding stellar magnitude.Taking each navigation star as the center,and superimposing multiple Gaussian distributions around it to form its standard identification image.The identification image of navigation star constructed by the stellar magnitude can effectively reflect the interaction between stars,and reduce the proportion of high-magnitude stars in the identification feature composition.As for the identification image deformation and the change of image gray value caused by missing detection of high-magnitude star points,the robust perceptual hash algorithm(robust phash)is introduced to extract the robust phash feature of the standard identification image.The simulation results show that when the Gaussian-Poisson noise parameters are 0.08 and 1/2 respectively,the identification success rate of proposed method is 88.65%.After the star image preprocessing and navigation star identification,the CNS obtains enough information to further estimate the position and attitude of the high-altitude aircraft.While,the star centroid positioning accuracy will directly affect the estimation accuracy of the CNS attitude and position navigation parameters.When the same traditional centroid positioning method is applied to different star sensors,the positioning accuracy of navigation star centroid is basically the same,but the corresponding angular resolution of centroid positioning error is significant different.It will affect the CNS attitude estimation accuracy.Besides,in the traditional INS/CNS integrated navigation system,the INS assists in transforming the CNS attitude navigation parameters to the navigation frame.It can’t guarantee the high accuracy of attitude navigation parameters.For this issue,a multi-sensor aided CNS attitude estimation method is proposed.In this method,the positioning error of navigation star centroid is taken as the state,and the gyroscope output is taken as the measurement.The non-linear relationship between the state and the measurement is used to correct the positioning error of navigation star centroid.The simulation results show that when the method is applied to star sensors with the same field-of-view but different resolutions,the pixel positioning accuracy of navigation star centroid increases with the decrease of image resolution,while the angular resolution of centroid positioning error is basically the same.The infrared Earth sensor is introduced to measure the aircraft elevation angle and azimuth angle relative to the celestial frame,so as to construct the transformation matrix between the celestial frame and the navigation frame for the CNS attitude navigation parameters.To improve the infrared Earth measurement accuracy,the proposed method first uses the specific force provided by the accelerometer to estimate the aircraft altitude angle and azimuth angle.Then,the weighted least square method is used to fuse the infrared Earth measurement and the specific force estimation measurement.The simulation results show that after the infrared Earth measurement error is corrected,the estimation accuracy of aircraft elevation angle and azimuth angle is increased to 4.0176 "and4.9907".The CNS can’t estimate the aircraft altitude information only based on the observation angle change of the navigation stars.Aiming at this issue,an aircraft altitude estimation method based on infrared Earth measurement is proposed.The method uses the infrared Earth measurement to determine the aircraft position vector relative to the celestial frame.Then,the aircraft altitude is estimated via the space triangle relationship between the aircraft position vector,the Earth center and the Earth edge relative to the celestial frame.The simulation results show that the maximum altitude estimation error under the proposed method is within0.4% of the aircraft altitude true value.Aiming at the problem that small stellar angular distance in a single field of view limits the positioning accuracy,a multi-star joint positioning method is proposed.Multiple candidate navigation stars are first selected from the SAO star catalog according to the identified navigation stars’ observation and astronomical information.Then,the candidate navigation stars’ pseudo-observation information is calculated to enhance the CNS positioning navigation parameters dimension.The simulation results show that when the number of candidate stars M is 5 and the angular distance υ between the candidate star and the main star is 103°,the aircraft position estimation accuracy is the best.Based on the above CNS attitude and position navigation parameters,a simulation verification system based on the INS/CNS deep integrated navigation model is established,the aircraft attitude and position navigation parameters are estimated to verify the navigation model performance.Under the simulation environment,the estimation errors of aircraft yaw,pitch,and roll angles of the traditional INS/CNS integrated navigation mode are 68.4803″,46.7713″,and 78.2840″,respectively.The aircraft attitude estimation accuracy of the INS/CNS deep integrated navigation model is increased to 19.2540″,13.2417″,and 26.4957″,respectively.The latitude and longitude estimation errors of the traditional INS/CNS integrated navigation mode are 52.0624 m and 63.6710 m,respectively.The position estimation accuracy under the improved model is 14.8877 m and 17.8506 m,respectively.The position estimation accuracy of the INS/CNS deep integrated navigation model is 9.4846 m and 13.0377 m,respectively. |
PDF Full Text Request