Orthogonal kernel projecting plane for radar HRRP recognition

It is known that the subspace method is effective for radar target recognition. Its key step is to find a suitable low-dimensional subspace. In this paper, a novel subspace method, namely orthogonal kernel projecting plane (OKPP), is proposed for radar target recognition using high-resolution range profile (HRRP). The goal of OKPP is to maximize the between-class distance while minimizing the within-class distance. By introducing an orthogonality constraint into the objective function, we obtain the orthogonal basis vectors of OKPP. Comparing with the conventional kernel-based subspace methods, such as kernel principal component analysis (KPCA) and kernel Fisher discriminant analysis (KFDA), the nonlinear features extracted by OKPP reduce redundancy and improve the target recognition performance. The explicit expressions of the basis vectors of OKPP can be solved without using singular value decomposition (SVD) process, and thus reduces the computation complexity. The experimental results using measured data show that the proposed method has an encouraging performance.