Multiview HRRPs Recognition Based on Dynamic Structure and Parameters Neural Network

High-Resolution Range Profile (HRRP) contains rich target features and is easy to obtain, making it important data in the field of radar aerial target identification. However, many current studies have only focused on single-view static networks, without making full use of the multi-view information of the target. Besides, the structure and parameters of the network are also fixed, which cannot learn the target features dynamically. Therefore, this paper proposes a multi-view network (Dynamic Structure and Parameters Neural Network, DSPNN) that can dynamically adjust its own structure and parameters according to the input samples. Firstly, the network realizes the dynamic adjustment of network depth and convolutional kernel parameters through the skip gate mechanism and dynamic convolutional kernel. Next, the weights of the extracted features are reassigned through the spatial attention mechanism. Finally, the fusion of multi-view features is accomplished by the multi-input feature fusion module, which improves the feature adaptive extraction and fusion capability of the network. Experimental results demonstrate that the proposed method has superior aerial target multi-view HRRPs classification performance over other compared methods.