Target Classification by Conventional Radar Based on Bispectrum and Deep CNN.

Due to the restriction of the low-resolution systems and the interference of background clutter and environmental noise in the exploration process, the traditional classification and recognition algorithms of conventional radar for aircraft targets have low accuracy and poor feature stability. To solve the above problems, this paper proposes to apply high-order cumulant spectrum and deep convolutional neural network (CNN) to feature the extraction and classification of aircraft target radar echoes. Firstly, analyze the high-order statistical characteristics of aircraft echoes, calculate their bispectrum, and then enhance the generated bispectrum dataset. Finally, use the augmented dataset to train and test the deep CNN, and obtain the final classification and recognition results. Experimental results show that the proposed method can accurately classify and identify multiple aircraft targets in the dataset, indicating that the bispectral features can better reflect the target characteristics, and the classification method combined with the deep learning model has good classification and identification performance and noise robustness. [ABSTRACT FROM AUTHOR]