Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Automated Coastline Extraction Using Edge Detection Algorithms

Conor O'Sullivan, Seamus Coveney, Xavier Monteys, Soumyabrata Dev

TL;DR

This study evaluates four edge-detection algorithms (Canny, Sobel, Scharr, Prewitt) for automatic coastline extraction from Sentinel-2/SWED imagery, using preprocessing steps (scaling, histogram equalization, Gaussian blur) and matching results to reference coastlines derived from labeled maps. Results show Canny achieves the best SSIM and closest visual edge alignment on average, but remains susceptible to development-induced noise; preprocessing improves edge discrimination, while some methods introduce spurious edges. The authors provide an open-source implementation and frame the work as a stepping stone toward deep-learning coastline segmentation, highlighting feature-engineering techniques that can benefit model training and robustness. Overall, the work offers a practical benchmarking of classical edge detectors on a diverse coastline dataset and informs preprocessing choices that can enhance automated shoreline monitoring.

Abstract

We analyse the effectiveness of edge detection algorithms for the purpose of automatically extracting coastlines from satellite images. Four algorithms - Canny, Sobel, Scharr and Prewitt are compared visually and using metrics. With an average SSIM of 0.8, Canny detected edges that were closest to the reference edges. However, the algorithm had difficulty distinguishing noisy edges, e.g. due to development, from coastline edges. In addition, histogram equalization and Gaussian blur were shown to improve the effectiveness of the edge detection algorithms by up to 1.5 and 1.6 times respectively.

Automated Coastline Extraction Using Edge Detection Algorithms

TL;DR

This study evaluates four edge-detection algorithms (Canny, Sobel, Scharr, Prewitt) for automatic coastline extraction from Sentinel-2/SWED imagery, using preprocessing steps (scaling, histogram equalization, Gaussian blur) and matching results to reference coastlines derived from labeled maps. Results show Canny achieves the best SSIM and closest visual edge alignment on average, but remains susceptible to development-induced noise; preprocessing improves edge discrimination, while some methods introduce spurious edges. The authors provide an open-source implementation and frame the work as a stepping stone toward deep-learning coastline segmentation, highlighting feature-engineering techniques that can benefit model training and robustness. Overall, the work offers a practical benchmarking of classical edge detectors on a diverse coastline dataset and informs preprocessing choices that can enhance automated shoreline monitoring.

Abstract

We analyse the effectiveness of edge detection algorithms for the purpose of automatically extracting coastlines from satellite images. Four algorithms - Canny, Sobel, Scharr and Prewitt are compared visually and using metrics. With an average SSIM of 0.8, Canny detected edges that were closest to the reference edges. However, the algorithm had difficulty distinguishing noisy edges, e.g. due to development, from coastline edges. In addition, histogram equalization and Gaussian blur were shown to improve the effectiveness of the edge detection algorithms by up to 1.5 and 1.6 times respectively.
Paper Structure (12 sections, 6 figures, 1 table)

This paper contains 12 sections, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related work
  3. Contributions of this paper
  4. Methodology
  5. Dataset
  6. Pre-processing
  7. Edge Detection and Metrics
  8. Results & Discussion
  9. Subjective visual analysis
  10. Objective evaluation metrics
  11. Effect of pre-processing
  12. Conclusion & Future Work

Figures (6)

  • Figure 1: Example of a test image, binary land/water map, and coastline edge obtained from SWED seale2022swed.
  • Figure 2: Example of pre-processing steps using blue band as an example.
  • Figure 3: We demonstrate the results obtained from edge detection algorithms applied to the blue band. We observe Canny to provide the best edge map followed by Prewitt.
  • Figure 4: We apply Canny edge detection to all channels of a SWED image.
  • Figure 5: Average PSNR for Canny edge detection with and without histogram equalization
  • ...and 1 more figures