基于GAN的雷达HRRP数据增强方法

在雷达自动目标识别(RATR)中,数据驱动方法是强有力的工具之一.然而数据驱动方法的性能十分依赖数据集的质量,数据增强方法通过扩充数据集,能够提升数据驱动模型在现有数据集上的识别率.本文提出了用于高分辨距离像(HRRP)数据生成的一维基础生成对抗网络(BGAN)结构和条件生成对抗网络(CGAN)结构,并利用生成的人工样...

Radar HRRP Data Enhancement Method Based on GAN

In Radar Automatic Target Recognition （RATR）， data-driven models have proven to be a powerful tool. However， the performance of the data-driven models were dependent on the quality of the data set. The data enhancement method could improve the recognition performance of the data-driven models on the existing data set by expanding the data set. This paper proposes a one-dimensional basic generative adversarial network （BGAN） structure and a conditional generative adversarial network （CGAN） structure for high resolution range profile （HRRP） data generation. Then using the generated artificial samples to complete the data enhancement. Experiments show that the two networks proposed in this paper can effectively improve the accuracy of target recognition， and the performance is better than the traditional translation and mirroring enhancement methods. The BGAN-based HRRP data enhancement method has the best performance， but its time and space complexity are relatively high； the CGAN-based data enhancement method can reduce the time and space complexity of the model while ensuring the increase in accuracy， and has high application prospects.