SAR Sea Clutter Data Generation Based On Improved Pix2pix Network

Sea clutter is an important factor for the detection of sea surface targets in radar images. However, only limited time and local sea clutter samples can be obtained currently, which cannot cover the ever-changing marine environment. The generation of sea clutter data from unknown sea areas based on complex marine environments has important theoretical significance and application value. Therefore, a sea clutter data generation method based on Pix2Pix is proposed for SAR images. The nonlinear mapping relationship between wave spectrum and SAR image spectrum is learned by 2912 pairs of SAR images and ERA-5 wave spectrum data. Thus, the network can generate matched sea clutter images by inputting real marine environment information (wave spectrum) after training. Finally, the image similarity index and histogram model fitting are used to verify the effectiveness of the proposed method.