Abstract: For broadband multi-polarization radar, this paper proposes an algorithm that combines high resolution range profile (HRRP) and polarization information to obtain the HRRP of target under four polarization configurations which is composed of the polarization distance matrix. This algorithm performs full-scale feature extraction and modeling of the target to adapt to different poses and effectively reduces the impact of HRRP azimuth sensitivity. Then, distance matrix, Pauli decomposition and Freeman decomposition are used to extract the target features of the polarization distance matrix, and the obtained target feature vectors are combined and sent to the constructed deep convolutional neural network for training and learning. This method not only combines different feature extraction methods to extract more comprehensive features of the polarization distance matrix, but also deeply learns the target eigenvectors by using the deep convolution neural network. The simulation results verify the effectiveness of the proposed method.