Abstract:

To extract the effective features of the high-resolution range profile (HRRP), a radar target recognition approach based on the extended feature of linear convolution coefficients is proposed. By this approach, the joint features which are composed of HRRP and its linear convolution coefficients are selected in the kernel space, and then the support vector machine (SVM) is applied to classification. Feature selection in the kernel space has the ability to solve the high dimensionality problems and nonlinear separability problems of joint features, thus improving the SVM recognition performances. On the other hand, the extended feature of linear convolution coefficients is more stable compared with HRRP. Moreover, this extended feature can make up for the losses of partial range cells caused by feature selection. Simulations results based on the HRRP dataset of five aircraft models demonstrate the validity of the proposed approach.