Automated Fish Cages Inventoryng and Monitoring Using H/A/α Unsupervised Wishart Classification in Sentinel 1 Dual Polarization Data

Aquaculture fish cages are usually located remote areas with poor accessibility. Synthetic Aperture Radar (SAR) is relevant to register and monitor their activity. However, in order to take advantage of the large amount of information this technology produce, it is necessary to develop automated tools of analysis. This work proposes an automated approach through an unsupervised polarimetric classification method using Sentinel 1 IW SLC Dual-Pol (VV+VH) products. An experimental evaluation was applied in Calbuco, an area with intense aquaculture activity in southern Chile. It was possible to demonstrate that this approach allows to improve the capacity of classification of previous experiences (86.49% user's accuracy and 96.97% producer's accuracy). Further studies are required to know the impact of wind speed on the classification as well as the spatial precision of the detection.