Abstract
This innovation note outlines an advanced approach for detecting and monitoring aquaculture ponds through deep learning and remote sensing techniques. Utilizing Sentinel-2 satellite imagery, the Normalized Difference Water Index (NDWI), and a combined DeepLabv3 and Random Forest classification model, it ensures precise identification and spatiotemporal tracking of aquaculture ponds. The methodology highlights the integration of shape parameters, spectral analysis, and automated processes to enhance accuracy and scalability. By supporting sustainable marine spatial planning, reducing environmental impacts, and promoting efficient resource management, this study contributes to coastal ecosystem conservation, sustainable aquaculture development, and informed policymaking.
