Aircraft Target Classification Based on CNN

In this paper, we applied the idea of deep learning to aircraft targets recognition based on time-frequency diagram. Firstly we introduced application of Convolutional Neural Network (CNN), and the methods of radar target recognition. Secondly, Short Time Fourier Transformation (STFT) was introduced. Thirdly, the structure of improved LeNet CNN was described, considering the character of radar echo wave. Fourthly, 4 kinds of aircraft targets were introduced. Then, the algorithm based on CNN and STFT was validated based on measured data, and was compared with Support Vector Machine (SVM). The accuracy rate could reaches up to 99.98%, 25% higher than SVM. Finally, we summarized advantages of the method proposed in this paper and give the suggestion in engineering application.