Preprocessing Technology Research Of Star Image Under Complex Working Condition

The high-precision measurement of spacecraft attitude is an important prerequisite for achieving high-precision attitude control of the spacecraft. Star sensor has the highest attitude accuracy among sensors broad to measure the attitude of spacecraft. Research concentrated on the preprocessing technology under complex operating conditions is of great significance. This article is divided into five main areas of study: study on high accuracy centroiding method of star point, pretreatment-technology research under high dynamic condition, star tracking algorithm design under high dynamic condition, star point extraction algorithm research under complex operating background and research on its implementation.By Fourier analysis, systematic error of CoM(Centre of Mass) and its modified form is studied on theoretical view, proving that systematic errors is cyclical in pixel size of the detector. Size of the centroiding pixel-extraction window and pixel size, center of the star and form of the window used to centroid the star are factors which cannot be neglected to influence the form of systematic error. A larger centroiding window used might help to eliminate the systematic error. Effect of noise on determining the centre of star is deeply studied, noise modeled as Gaussian white noise and Poisson distribution are considered, the CRLB(Cramer lower boundary) of both model are deduced. By numerical analysis, it is proved that Gamma correction can improve the accuracy of star centre in low SNR(Signal to noise ratio), while precision of CoM is mainly limited by the systematic error in high SNR.From the optical model of the star sensor, trajectory of the star centre in the detector plane that driven by three-axis angular velocities is deduced. Combining the star imaging model under static condition, following a Gaussian distribution, star imaging model in high dynamic condition is given. Under high dynamic condition,drives the centre of star sliding along the y axis direction, drives the centre of star sliding along the x axis direction and drives the centre of star sliding along an arc. Integration time of the star sensor can be adjusted with the variation of three-axis angular velocities to ensure the centre of the star to move a distance that is nearly as long as a star spot width, PS(Peak signal) of the star would reach its maximum value, reducing the dynamic trailing of the star simultaneously.Motion model of star in detector plane of star sensor is established. To meet demand of tracking model under high dynamic condition, motion model which considering the acceleration and jerk information of the star is reconstructed for high accurate prediction of the track window position under dynamic conditions, based on the differential star tracking method. Not depending on prior pose information, predictor of coordinate of tracking window is proposed based on the extended Kalman filter. Simulation results show that the proposed algorithm can work well even if the angular velocity reaches 10deg/s. The outdoor star test shows that our method keeps high accuracy which is about less than 5 pixels when the angular velocity reaches 4.0deg/s at least.Characters of star image in the Moon view and high dynamic condition is studied. Star gray extraction algorithm based on the gray distribution features of star point is proposed. Compared with the global threshold method and local adaptive threshold method, the proposed method can only scan the star pixel once time and then centre of the star point is calculated. Also, bilateral threshold is used to overcome the complex variation of background caused by strong radiation, using regional growth strategy to extract more star pixels that are out the extraction window. According to the place of seed point, gray value of star pixel and its coordinate are multiplied. And accumulate operation is finished at the same time.Take advantage of the parallel processing features of FPGA, the proposed star-extraction algorithm under complex operating condition is implemented in hardware. Extraction of star pixel, estimation of target area and target energy, add and multiply-accumulation operation of star gray value and corresponding coordinate and the output of image pixel is operated at the same time. Centroiding is just after a frame of star image is finished, as is in real time meeting the demand of preprocessing algorithm of star sensor.