Enhancing Water Depth Inversion Accuracy via SAR and Variable Window Sliding Segmentation

Meng Zhang ^1* Chao Qi ¹ Fanlin Yang ¹ Ruifu Wang ¹ Saeid Pirasteh ²

• ¹ Shandong University of Science and Technology, Qingdao, China
• ² Institute of Artificial Intelligence, Shaoxing University, Shaoxing, China

The utilization of synthetic aperture radar (SAR) for depth inversion is crucial for accurate underwater mapping. However, current SAR-based techniques face challenges in segmentation accuracy, which directly affects inversion precision and spatial resolution. Traditional segmentation methods lack efficiency and often result in low-resolution outcomes. To address these issues, we propose a novel SAR water depth inversion method based on variable window sliding segmentation. This method optimizes nearshore image utilization by dynamically adjusting the pixel size and preventing coastline encroachment, leading to more precise swell wavelength measurements. When applied to the eastern sea off Naraha, Japan, our method achieved a minimum mean relative error (MRE) of 9.2% for shallow waters (0 to 20 m depth) and 4.9% for deeper waters (80 to 100 m depth). These results significantly improve upon those of traditional methods, which typically show MREs ranging from 10% to 30%. Additionally, our method achieves a maximum spatial resolution of 5.5 m, a notable advancement in nearshore depth measurement.The study also revealed that different depth ranges and function types, particularly linear and atanh functions, impact measurement performance, demonstrating superior accuracy across multiple metrics.