Ship HRRP Target Recognition Based on CNN and ELM

Abstract

A new network structure CNN-ELM combining convolutional neural network (CNN) with extreme learning machine (ELM) is proposed in this paper. It is applied to ship high resolution range profile (HRRP) target recognition and implements automatic extraction of deep features of data. First, the CNN-ELM model is constructed and the network parameters are set. Second, in terms of the problem that HRRP data is one-dimensional, HRRP data are reordered to transform one-dimensional data into two-dimensional data. Then, the CNN-ELM model with training data is trained to obtain network parameters. Last, the recognition of test data with trained network model is targeted. In the experiment, the recognition rate of CNN-ELM reached 99.50%. In order to verify the generalization performance of the new method, the data set were halved and CNN-ELM was compared with different methods. The recognition rate of CNN-ELM reached 91.13% in the case of halving the data set. The experimental results of the measured data show that CNN-ELM has a better recognition performance.