Research On Key Technologies Of High-accuracy Attitude Measurement By Star Sensor

Star sensors are attitude sensors characterized by high accuracy,low power consumption,long life,no drift error,strong anti-interference,and independent navigation without reliance on other systems.Currently,star sensors are widely used in the aeronautics and aerospace fields.Moreover,they have become the mainstream direction of the attitude sensor-related research.At present,star sensors are high-accuracy instruments used for flight vehicle attitude determination.The accuracy is the most important index of star sensors,since it directly affects the overall performance of flight vehicles.The complete accuracy evaluation theory of star sensors is the basis for the development of high-accuracy star sensors.In addition,it indicates the research direction as regards the key technologies for the high-precision attitude determination of star sensors.First,research on the attitude measurement accuracy evaluation is conducted.The conventional precision analysis of a star sensor cannot reflect its current status in real-time.Therefore,this paper proposes a real-time attitude precision analysis method for a star sensor.An attitude measurement model of a star sensor is developed based on a pinhole imaging model.Combined with the error transfer formula of over-determined equations for the multivariable implicit function proposed in this paper,the error transfer model is analyzed from three different aspects: random errors of starspot extraction,random errors of star catalog,and systematical errors of intrinsic parameters.On the one hand,the real-time accuracy analysis of star sensors proposed in this paper is validated through simulation;on the other hand,the analysis reveals that the on-orbit calibration of intrinsic parameters,starspot detection in complex background,and star identification are the key factors affecting the high-accuracy attitude determination of star sensors.Subsequently,this paper investigates the above three key factors of high-precision attitude determination one by one.In the on-orbit calibration of intrinsic parameters,the high-order distortion compensation is very important for high-precision attitude determination.For the star sensor attitude-independent on-orbit calibration,the coordinates of starspots without distortion on the image plane are not the most significant parameters.In addition,the direct correction of high-order distortion requires a large amount of star data.This is due to that the mutual effect of the high-order distortion parameters makes it difficult to obtain stable results when a small amount of training data is employed.Since the coordinate estimation of starspots without image plane distortion is unnecessary,the focal length polynomial can be utilized for high-order distortion correction.To this end,a star sensor attitude-independent on-orbit calibration method is proposed.First,the zero-order distortion is compensated using the coordinate offsets;second,the higher-order distortion terms are merged as a part of the focal length function;finally,the order of the focal length polynomial is determined through cross-validation.Due to factors such as nonuniform response,stray light,and cloud or skyline interference,starspot detection has become a difficult task in the design of star sensors operating in complex light environments.To detect small targets such as starspots under dense clutter conditions,a single-frame target detection algorithm based on a small bounding-box filter is proposed.This algorithm is characterized by good adaptability to both the position and size of a small target.During the detection process of small targets,the proposed algorithm first searches for the local maximum gray pixel.Subsequently,a set of concentric bounding boxes whose center is the pixel found in the previous step is constructed.Next,based on the bounding boxes,detection thresholds of a neighboring region of this pixel are calculated.In the end,the minimum threshold is used to detect small-target pixels in the neighboring region.A fast version of the proposed algorithm is a minimum bounding-box filter.This can be implemented by dividing an image into blocks and using the mid-range as well as range to assess the concentration trend and dispersion of the background.Since the initial establishment of the attitude requires a high star-identification rate,conventional star sensors have rigorous requirements regarding the number of stars to be detected.In the pre-processing approach with the detect-before-track method,a contradiction between the low false alarm rate of noise and the high detection probability of low signal-to-noise-ratio(SNR)star identification.Consequently,this paper proposes a star sensor working mode,where a track-before-detect approach is employed.Furthermore,since the star identification with a small number of starspots is not unique,this work proposes a star identification method based on multiple hypotheses testing.The proposed method generates a star match tree through star trajectories,and gradually prunes the star match tree to achieve star identification.Finally,the experimental and analysis results indicate that the proposed method can achieve a better calibration effect using fewer star data than traditional attitude-independent on-orbit calibration methods.In particular,when the star sensor has complex distortion,the proposed method can reduce the amount of required star data by approximately 50% compared to traditional methods.Moreover,the small bounding-box filter can achieve high detection probability and low false alarm rates when detecting small targets in dense clutters.In addition,the fast version of the proposed algorithm is characterized by high computational efficiency.The star identification approach with multiple hypotheses testing can be completed when the starspot extraction error reaches 0.5 pixels,SNR=3.8 and only a small number of guide stars is detected.The main contributions of this paper are as follows: a)Through the moving frame,the defects of the Euler angle representation of the attitude in the accuracy analysis of star sensors are overcome.An error transfer model of multivariable implicit over-determined equations is proposed,which improves the error analysis theory.Furthermore,the attitude accuracy of the star sensor is evaluated in real-time based on the Euler angle under the moving frame;b)Regarding the on-orbit calibration of intrinsic parameters,a method of using a focal length polynomial to compensate the high-order distortion of the star sensor optical system is proposed,which overcomes the coupling of the distortion coefficient and can realize high-order distortion correction with a small amount of guide star data;c)A single-frame starspot target detection method with a small bounding box is proposed.The filter possesses the characteristics of adaptive position and size,and its fast version has high operation efficiency;d)Based on a track-before-detect approach,a star identification method with multiple hypotheses testing is introduced.