A new method of high resolution radar target recognition based on Convolution Neural Network (CNN) was presented. To solve the problem of slow convergence of loss function values during the training process when small samples are applied to the deep CNN, High Resolution Range Profile (HRRP) features were firstly extracted by using the improved CNN combined with the Batch Normalization (BN) algorithm, and then classified by using a Support Vector Machine (SVM). The experimental results using high-fidelity electromagnetic simulation data of military vehicles validate the effectiveness of the proposed method.