Research on Supervised Manifold Learning for SAR target classification
Juan Wang, Lijie Sun · 2009
Nonlinear manifold learning algorithms, mainly isometric feature mapping (Isomap) and local linear embedding (LLE), determine the low-dimensional embedding of the original high dimensional data by finding the geometric distances between samples. This paper proposed an approach to reduce the dimensions of SAR image targets based on supervised manifold learning algorithm. Three steps were done to reduce the dimensions of original data. Firstly take use of a priori information of the sample point to find the k-neighbors. Secondly to build the local reconstruction weight matrix W. Thirdly get the dimension reduction result based on W and the neighborhood of original data. Experiments were done to test the effect of dimensionality reduction to classification results. Three types of targets were used in the experiments. The implementation steps and parameter settings are discussed in details. The results show SLLE is more conducive to SAR image target classification than the traditional LLE.