Automatic Census of Mussel Platforms Using Sentinel 2 Images

TL;DR

This paper presents an automated method using Sentinel 2 satellite images, remote sensing indices, and neural networks to census and monitor mussel platforms in Galician estuaries.

Contribution

It introduces a novel approach combining multispectral satellite imagery and neural networks for automatic census of floating aquaculture structures.

Findings

Successful detection of mussel platforms in satellite images

Effective use of remote sensing indices and neural networks

Potential for regular monitoring of aquaculture sites

Abstract

Mussel platforms are big floating structures made of wood (size is normally about 20x20 meters or even a bit larger) that are used for aquaculture, id EST: growing mussels in appropriate marine waters. These structures are very typical in Galician estuaries. Being interesting to produce a periodic census of these structures that would allow knowing their number and positions, as well as changes on those parameters; Satellites that obtain periodic images for earth observation are a natural election for this issue. This paper describes a preliminary application able to construct automatically such a census using Sentinel 2 images (Copernicus Project). Copernicus satellites are run by European Space Agency (ESA) and the produced images are freely distributed on the internet. Sentinel 2 images have thirteen frequency bands and are updated each five days. In our application, we use remote sensing normalized (differential) indexes and artificial Neural Networks applied to multiband data. Different methods are described and tested. Finally, results are presented.