Pre-processing Of Star Sensor Image Denoising And Hardware Implementation

Of all current methods for measuring spacecraft attitude,the use of star sensor produces the most accurate measurements.Compared with other sensors,star sensor has the advantages of autonomy and no attitude error accumulation,making it possible to use a gyro-less attitude determination system on spacecraft.The extraction of star coordinate is one of the key technologies in star sensor research and development,and the extraction precision of star coordinate and real time are important criterions.The star image filtering algorithm is investigated in this dissertation,starting with how to reduce the star image noise and improve the extraction accuracy of star coordinate.The main contents of this dissertation are as follows:Expounding the working principle and hardware structure of the star sensor and analyzing the precision of the attitude measurement of the star sensor.Explaining main accuracy indexes of star sensor,including attitude angle measurement accuracy,noise equivalent angle,optical axis pointing accuracy and single star measurement accuracy.Analyzing the factors affecting the measurement accuracy of star sensor,including the limitation of pixel size and the error caused by measurement,for the latter,mainly divided into three categories,optical system error,algorithm error and error caused by the noise.For the error above,methods to improve the accuracy of star sensor are described,including advanced calibration techniques,optimization algorithm,noise suppression and preprocessing of noise reduction of star maps.After contrasting a few filter types,this paper choose s two-dimensional FIR filter as the processing of the star graph,which is more suitable for the star image filtering.The 25(5 ? 5)order two-dimensional FIR filter was designed by using window function method and frequency sampling method,the obtained filter parameters could be programmed by host computer to filter the star maps,and then the spectrum of the star images before and after filtering are analyzed.Based on the star maps processed by the filter designed by the two methods above,frequency analysis is achieved and effects of filtering are contrasted.To investigate the influence of filter order,frequency sampling method is adopted to design a 9(3 ? 3)order two-dimensional FIR filter,the effect of which is compared to the effect of the 25(5 ? 5)order two-dimension FIR filter,so that the conclusion is drawn.According to theoretical analysis and simulation experiment,this paper chooses the parameters of 25(5 ? 5)order two-dimensional FIR filter designed by frequency sampling method,which is programmed in VHDL language.And then the filter module successfully adjusted is integrated to system to process the preprocessing of noise reduction of star map.With the coordinates of star obtained by FPGA uploaded to host computer,the results of which is in contrast with the results of coordinate extraction that is not included in the FIR filter module.The experiment result shows that star map filtering could make the effect of high frequency noise suppression remarkable,and improve the accuracy of coordinate extraction of fixed star.