Multiscale Kernel Sparse Coding-Based Classifier for Hrrp Radar Target Recognition

With the combined multiscale Gaussian kernel and Morlet wavelet kernel, two multiscale kernel sparse coding-based classifiers (MKSCCs) are proposed for radar target recognition using high-resolution range profiles (HRRPs). The kernel trick can make samples more clustered in higher-dimensional space. Moreover, the multiscale kernels at different scales have advantages of good generalisation and primary signature capturing ability for target's HRRP, which are helpful to improve the target recognition accuracy and robustness of MKSCC further. Numerous experiments are conducted on five types of ground vehicles' HRRP data and the authors also make comparisons with the KSCC and some related recognition methods. The results demonstrate the effectiveness of the proposed method.