| Satellite attitude measurement is the fundament of satellite attitude control, which plays an essential role to the satellite normal running. Star sensor is an attitude sensitive measurement system with high-accuracy. The key to attitude measurement is how to realize star pattern recognition using star sensor. Star pattern recognition is to confirm the corresponding relations between the star map of observed stars in field of view (FOV) of star sensor and the star map of guide stars.Star pattern recognition can be considered to be a kind of pattern recognition. Consequently, we have proposed a star pattern recognition method of star sensor based on the neural network technology, a recognition system has been designed with containing parallel BP multi-subnet. The construction of input and output sample, designing of classifier and designing of the training network configuration are introduced respectively, some improvement algorithms of BP neural network are discussed. The efficiency of star pattern recognition greatly is improved through ameliorating configuration of network and program method.The more satisfactory result is obtained through simulating the star pattern recognition process on the simulation platform. The result indicates that it is feasible to make use of neural network technology to realize star pattern recognition, and it is easy to be realized. This algorithm effectively overcomes the contradictions of traditional algorithm among recognition success rate, recognition time and storage capacity. And this algorithm has been improved notably in aspects of database capacity and real-time, comparing with traditional star pattern recognition algorithm.
|
PDF Full Text Request