Research On Software Algorithm Of Star Sensor Based On SOM

The navigation technology of an aircraft is one of the key technologies for space exploration,and the acquisition of aircraft attitude is the basics of navigation technology.At present,various attitude sensors have been developed,such as,earth sensor,sun sensor,star sensor,etc.Because of the high measurement accuracy,strong anti-interference ability and the capacity to realize autonomous navigation,star sensor has become one of the most important attitude measurement instruments,which can also be applied to missiles and airplane.Attitude update rate,recognition rate and location accuracy are important indicators of star sensor,besides that,star centroid location,star recognition and attitude calculation are the key algorithms of star sensor.Because the star recognition directly affects the recognition rate and time of star sensor,and thus affects the attitude update rate,it is of great significance in researching.Therefore,this paper will focus on the star recognition algorithm and propose a new recognition method based on SOM network and triangle algorithm.Although other researchers had proposed many recognition methods,there are still existing problems such as low feature dimensions and many redundant matching.Neural network technology is produced with the development of artificial intelligence technology.It has special capabilities of learning and fault tolerance,so it has been applied to many fields and achieved excellent results.In this paper,the neural network intelligent algorithm is applied to the star recognition,using the SOM network to classify stars into multiple classes,and then applying triangle algorithm in the corresponding star catalogue to find the matched triangle.In the whole sky identification,the recognition rate is still as high as 99% on the simulated star map with mean Gaussian noise of 0.1 and standard deviation of 0.025.Star centroid location is the base of the star recognition algorithm,and the positioning accuracy indirectly affects the recognition speed and accuracy.So this paper will also explore each step of the star centroid extraction: select common algorithms to apply and compare the advantages and disadvantages of each algorithm.Because the research on software algorithm of star sensor is closely related to the star map,and it's hard to get the measured star maps,so many researchers experiment with simulated star map instead of the measured map.This paper uses a large number of simulated star maps to compare the centroid extraction algorithms and verify the recognition algorithm in the early research of algorithm.When the algorithm is well-debugged,we apply the algorithm to the measured star maps and compare the application effects on different maps.In addition,in this the paper,the practicability of the recognition algorithm on the SLR camera has been verified,thus provides a way for the functional verification of the recognition algorithm.