基于自注意力机制和CycleGAN的高分三号ScanSAR图像的扇贝效应抑制

Descalloping of GF-3 ScanSAR image based on self-attention mechanism and CycleGAN

高分三号卫星是我国首颗分辨率达到1 m的C波段多极化合成孔径雷达(synthetic aperture radar, SAR)卫星,其中扫描式合成孔径雷达(scan synthetic aperture radar, ScanSAR)模式是高分三号卫星重要的工作模式之一,由于该模式的工作机制导致生成的图像可能发生扇贝效应,一般呈现为明暗相间的条纹。本文针对高分三号卫星ScanSAR模式下存在的扇贝效应,提出自注意力机制与循环一致对抗网络(cycle-consistent adversarial networks, CycleGAN)结合的模型对ScanSAR图像进行处理,从而抑制扇贝效应产生的条纹现象。本文所示方法与传统扇贝效应抑制方法和深度学习相关算法进行比较,并通过亮度均值、平均梯度等指标进行分析。实验结果表明,本文方法可以对高分三号ScanSAR图像存在的扇贝效应进行较好的处理,有效抑制图像的条纹现象,使得图像质量得到提升,具有较大的实用意义。

GF-3 satellite is the first C-band multi-polarimetric synthetic aperture radar satellite with a space resolution up to 1 m in China,in which scan synthetic aperture radar(ScanSAR)is one of the important mode of GF-3.The working mechanism of this mode results in the phenomena of serious nonuniformity,generally showing bright and dark stripes,also known as scalloping.In view of scalloping in ScanSAR mode of GF-3,this paper proposes a model combining self-attention mechanism and cycle-consistent adversarial networks(CycleGAN),so as to perform descalloping.The proposed descalloping method is compared with traditional descalloping methods and deep learning related algorithms,and analyzed by indicators such as brightness average and average gradient.The experimental results demonstrate that the proposed method in this paper can better complete descalloping in the GF-3 ScanSAR image,effectively suppress the stripes phenomenon of the image,and improve the image quality,which is of great practical significance.