Radar target recognition using kernel uncorrelated discriminant local tangent space alignment

In this paper, a new manifold learning algorithm called kernel uncorrelated discriminant local tangent space alignment (KUDLTSA), is proposed for radar target recognition. The aim of KUDLTSA is to preserve the intrinsic geometric structure of a data set, which is represented by the local tangent space, while maximizing the between-class distances. Moreover, a simple uncorrelated constraint is introduced to get statistically uncorrelated features, which is helpful to improve the performance of radar target recognition. Optimizing the objective function in the reproducing kernel Hilbert space, nonlinear features are extracted. Experimental results on radar target recognition with high range resolution profiles demonstrate the effectiveness of the proposed algorithm.