A Convolutional Neural Network for Coastal Aquaculture Extraction from High-Resolution Remote Sensing Imagery

Aquaculture has important economic and environmental benefits. With the development of remote sensing deep learning technology, coastline aquaculture extraction achieved rapid, automated, high-precision production. However, some problems still exist in extracting large-scale based on high-resolution images: (1) generalization models caused by diversity breeding areas; (2) confusion target identification complex background interference land sea; (3) boundary area is difficult to extract accurately. In this paper, we built a comprehensive sample database spatial distribution aquaculture, expanded using confusing objects as negative samples. A multi-scale-fusion superpixel segmentation optimization module designed solve problem inaccurate boundaries, coastal network proposed. Based dataset that labelled produced ourselves, extracted cage culture areas raft near mainland China images. The overall accuracy reached 94.64% state-of-the-art performance.