Enhancing coastal water body segmentation with Landsat Irish Coastal Segmentation (LICS) dataset

TL;DR

The dataset is used to evaluate various automated approaches for segmentation, with U-NET achieving the highest accuracy of 95.0% among deep learning methods, Nevertheless, the Normalised Difference Water Index (NDWI) benchmark outperformed U-NET with an average accuracy of 97.2%.

Abstract

Ireland's coastline, a critical and dynamic resource, is facing challenges such as erosion, sedimentation, and human activities. Monitoring these changes is a complex task we approach using a combination of satellite imagery and deep learning methods. However, limited research exists in this area, particularly for Ireland. This paper presents the Landsat Irish Coastal Segmentation (LICS) dataset, which aims to facilitate the development of deep learning methods for coastal water body segmentation while addressing modelling challenges specific to Irish meteorology and coastal types. The dataset is used to evaluate various automated approaches for segmentation, with U-NET achieving the highest accuracy of 95.0% among deep learning methods. Nevertheless, the Normalised Difference Water Index (NDWI) benchmark outperformed U-NET with an average accuracy of 97.2%. The study suggests that deep learning approaches can be further improved with more accurate training data and by considering alternative measurements of erosion. The LICS dataset and code are freely available to support reproducible research and further advancements in coastal monitoring efforts.

Figures (12)

• Figure 1: Summary of the Landsat scene selection, scene cropping and annotation process. The end result of the process is 30,000 training instances and 100 test instances.
• Figure 2: An example of each of the 11 Landsat tiles considered for this analysis. The tile's row and path (row,path) are given in the title above each image. The scenes have been visualised using the NIR band to show contrast between land and ocean.
• Figure 3: Frequency of cloud cover percentage. The frequencies are calculated using the metadata of 14,850 Landsat scenes of Ireland.
• Figure 4: Average solar altitude by month of 14,850 Landsat scenes of Ireland. We take the altitude of the sun at the location and time the scenes were taken. We can see that the altitude is highest in the summer months.
• Figure 5: Example of test and training crops from a Landsat scene with tile (205,23). The test crop is given by the red square. The training crops are shown by the 300 blue squares.