Satellite Target Recognition Algorithm Based on BP Neural Networks

For high resolution range profile (HRRP) is sensitive to pose and translation, back-propogation (BP) algorithm is proposed to be used to process even rank central moments of HRRP in target recognition. Wavelet denoising is used to enhance the signal noise rate (SNR) of HRRP. Then central moments are extracted from the denoised HRRP. Even rank central moments can be used as features for target recognition because they are more stable and the dimension is reduced. BP algorithm is used to process the central moments feature vector. The experimental results based on real satellites data show that the proposed method achieves good recognition performance based on its low storage and computational complexity.