The neural network based on attention mechanism can focus on extracting the feature information from the key areas. The application of this characteristic in the polarimetric imaging target classification can help us to obtain the relationships among different polarimetric images and to extract more feature information from critical areas. To solve the difficulty of target recognition in cluttered natural backgrounds, this paper presents a polarimetric imaging target classification method based on an attention mechanism. Firstly, the attention mechanism and the convolutional neural network are combined to construct a polarimetric feature extraction model suitable for limited samples. Then, proper polarimetric images are selected as the input model for training so that the attention module can give more weights to the channel domain and spatial domain feature information that is easily classified to obtain higher classification accuracy. The experimental results show that the classification accuracy of the proposed method can be further improved in different natural backgrounds and reach more than 95% in the self-built polarimetric target database, which is obviously improved to compare with that of the traditional deep learning classification method. Thus, our method is more suitable for target classification in cluttered backgrounds.