Radar HRRP Target Recognition Based on Composite Deep Networks

Radar target recognition is to extract the acquired target echo information to achieve the determination of target category and attribute. The feature extraction and classifier in radar target recognition determine the quality of the recognition. However, the shallow structure used by traditional feature extraction algorithms and classifiers cannot fully utilize the original measurement data of radar targets, and it is easy to ignore the hidden information in the data. To solve this problem, a target recognition method based on composite deep networks is proposed. In this paper, autoencoder and convolutional neural network are combined to make full use of the complex function representation capabilities of deep network structure. The effectiveness of the proposed method is demonstrated by experiments with radar high-resolution range profile (HRRP) data based on dynamic scene simulation. The experimental results show that the proposed method of radar target recognition can achieve better performance than traditional method.