Space target recognition based on 2-D wavelet transformation and KPCA

Space target recognition is an important topic to a country´s space safety. Based on analyzing the characteristic of target space images, and combined with Discrete Wavelet Transformation, Singular Value Decomposition and Kernel Principal Component Analysis, a new method for space target recognition is proposed. Firstly, the detail sub-images are obtained by two-dimensional DWT of the original space target images. Then, the singular value feature vector is extracted via the SVD of sub-images, and it is mapped onto the principal feature space with KPCA to obtain the nonlinear feature. Finally, space target recognition is realized according to K-Nearest Neighbors classifier. The experimental results on space target images prove that the recognition rate of the algorithm adopted in this study is higher than that of other algorithms, such as SVD and DWT-SVD.