A Novel Algorithm for HRRP Target Recognition Based on CNN

Compared with traditional methods, deep neural networks can extract deep information of targets from different aspects in range resolution profile (HRRP) radar automatic target recognition (RATR). This paper proposes a new convolutional neural network (CNN) for target recognition based on the full consideration of the characteristics (time-shift sensitivity, target-aspect sensitivity and large redundancy) of radar HRRP data. Using a convolutional layer with the large convolution kernel, large stride, and large grid size max-pooling, the author built a streamlined network, which can get better classification accuracy than other methods. At the same time, in order to make the network more robust, the author uses the center loss function to correct the softmax loss function. The experimental results show that we have obtained a smaller feature within the class and the classification accuracy is also improved.