Research On Orientation Measurement Of Ground Objects For Satellite Autonomous Navigation

With the continuous exploration of space,the number of space missions gradually increased,and the tasks become more complex.Realization of satellite autonomous navigation is the basis of satellites to complete a variety of space missions.Celestial navigation is the most commonly used method of satellite autonomous navigation,but it has low accuracy in the horizon direction by using infrared horizon method.In this article,a satellite autonomous navigation system based on scene matching and astronomical navigation is proposed to solve the shortcomings of astronomical navigation.We focus on the method of obtaining the direction vector to the ground object by using the scene matching method and the screening method of the reference image based on the Support Vector Machine.Firstly,an autonomous navigation system based on scene-matching and celestial navigation is proposed.The orientation vector is obtained by scene matching method.Combining the direction of starlight vector obtained by star-sensor in celestial navigation,the angle between two vectors is used to be observed quantity.The UKF is chose to realize satellite autonomous navigation.In the traditional astronomical navigation system,the method of obtaining the horizon direction based on the infrared horizon has a low precision.So the article make a change by using the method of image matching to obtain the horizon direction.According to the motion characteristics of the satellite,the real-time requirement of the satellite autonomous navigation system is far less than the requirement of matching precision.We choose the matching method based on feature which has better robustness than other matching methods.SIFT(Scale Invariant Feature)and SURF(Fast Robust Feature)are usual image features to describe images.Comparing two features and selecting the SIFT which higher accuracy and better robustness as matching feature for matching navigation.And fast search algorithm based on KD-tree and BFF is used to enhance the speed of feature matching.The RANSAC method is used to reduce the mismatch point and further improve the matching precision.The SIFT feature used in the scene matching system has good robustness.In view of this,measurement parameters we used for matching area selection should be jamming resistance.We select the correlation peak feature,the image variance,the improved phase independent pixel number,the effective contour density and the matrix singular value decomposition ratio as the measurement parameters.Based on SVM(Support Vector Machine),the measurement parameters can achieve the benchmark images.Experimental results show that this method has good adaptability and anti-jamming,and it can judge the correct decision for matching area selection at the accuracy of 86%.Finally,using the remote sensing images from Google Earth as experimental data,the orbit measurement simulation of the integrated navigation system is carried out based on the simulated orbital parameters generated by the STK.The feasibility of the system is verified by the simulation.And the results of the comparison of different filtering methods and the number of observations are analyzed in the simulation.