Radar target recognition based on fuzzy optimal transformation using high-resolution range profile

In this paper, a novel fuzzy optimal transformation (FOT) method for radar target recognition using high-resolution range profile (HRRP) is proposed. The goal of this method is to maximize the between-class distance, while preserving the within-class structure. Firstly, FOT selects a low-dimensional vector for each class as the congregating center of its subprofiles, and maximizes the between-class distance by increasing the separation between the pairs of those congregating centers. Then, FOT attempts to seek the fuzzy optimal transformation that can preserve the local within-class structure, and considers the samples contribution to each class by introducing the membership values when preserving the local structure. Thus, FOT has more discriminant power than the other methods. Moreover, in the sensing of optimal separation, we derive a set of equations for selecting optimal congregating centers. Experimental results on three airplane targets show that the proposed method not only provide better recognition performance, but also more robust in noise than the conventional subspace methods, such as Eigen subspace method and canonical subspace method.

► The between-class distance is Maximized based on optimal congregating centers.
► The within-class distance is Minimized using fuzzy optimal transformation.
► We derive a set of equations for selecting optimal congregating centers.