Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Vidhu Arora, Shreyan Gupta, Ananthakrishna Kudupu, Aditya Priyadarshi, Aswathi Mundayatt, Jaya Sreevalsan-Nair

TL;DR

CCESAR addresses coastline extraction from SAR imagery by introducing a two-stage workflow that first classifies coastline type (natural vs built) with a CNN and then segments using type-specific U-Nets. The approach is validated on Sentinel-1 data across 8-bit and 32-bit GeoTIFF formats, with ground-truth masks generated from OpenStreetMap polygons. Results show that CCESAR generally outperforms a single U-Net, especially for higher-precision 32-bit data, and demonstrates improved generalization across coastline types and compression levels. This work enhances robustness of SAR-based coastal monitoring by incorporating coastline-type information into the segmentation process and provides a dataset with mixed coastline types for broader application.

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.

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

TL;DR

CCESAR addresses coastline extraction from SAR imagery by introducing a two-stage workflow that first classifies coastline type (natural vs built) with a CNN and then segments using type-specific U-Nets. The approach is validated on Sentinel-1 data across 8-bit and 32-bit GeoTIFF formats, with ground-truth masks generated from OpenStreetMap polygons. Results show that CCESAR generally outperforms a single U-Net, especially for higher-precision 32-bit data, and demonstrates improved generalization across coastline types and compression levels. This work enhances robustness of SAR-based coastal monitoring by incorporating coastline-type information into the segmentation process and provides a dataset with mixed coastline types for broader application.

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.
Paper Structure (14 sections, 4 figures, 2 tables)

This paper contains 14 sections, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Dataset Preparation
  3. Data Collection
  4. CCESAR: Our Proposed Workflow
  5. Preprocessing of SAR Images
  6. Ground Truth Mask Generation
  7. Coastline Classification Model
  8. Coastline Extraction Model
  9. Model Training and Optimization
  10. Post-Processing and Model Evaluation
  11. Experiments, Results, and Discussion
  12. Experimental Setup
  13. Results
  14. Conclusions

Figures (4)

  • Figure 1: Two stages in our proposed model in CCESAR.
  • Figure 2: Our proposed two-stage deep learning model for image classification and coastline segmentation in CCESAR.
  • Figure 3: Results from implementing CCESAR on both natural and built coastlines with coastlines extracted from Sentinel-1 SAR images with (left) 32-bit and (right) 8-bit compression.
  • Figure 4: Examples of Sentinel-1 SAR images where CCESAR gives faulty results.