CCESAR: Coastline Classification-Extraction From SAR Images Using CNN-U-Net Combination

2025-01-21

AI Summary

Addressing the challenges of representing morphologically diverse coastlines and limited segmentation accuracy of single-model approaches in Synthetic Aperture Radar (SAR) imagery, this paper proposes CCESAR—a two-stage framework. In the first stage, a Convolutional Neural Network (CNN) performs coarse-grained classification of coastline types; in the second stage, a U-Net is guided by the classification output to achieve type-aware fine-grained segmentation. This classification-guided extraction paradigm is the first to introduce explicit type-awareness into SAR-based coastline extraction, significantly enhancing modeling capability for complex coastal structures. Evaluated on real Sentinel-1 SAR data, CCESAR is end-to-end trainable and demonstrates robustness across varying image compression levels. Compared to a baseline U-Net, CCESAR achieves simultaneous improvements in both coastline-type classification accuracy and segmentation Intersection-over-Union (IoU), validating its superiority and generalization capacity.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

In this article, we improve the deep learning solution for coastline extraction from Synthetic Aperture Radar (SAR) images by proposing a two-stage model involving image classification followed by segmentation. We hypothesize that a single segmentation model usually used for coastline detection is insufficient to characterize different coastline types. We demonstrate that the need for a two-stage workflow prevails through different compression levels of these images. Our results from experiments using a combination of CNN and U-Net models on Sentinel-1 images show that the two-stage workflow, coastline classification-extraction from SAR images (CCESAR) outperforms a single U-Net segmentation model.

Problem

• Synthetic Aperture Radar (SAR) Imagery
• Coastline Detection
• Classification

Innovation

• CCESAR
• CNN-U-Net
• SAR Image Processing