Optical Remote Sensing Image Waters Extraction Technology Based on Deep Learning Context-Unet

摘要

The optical remote sensing images are used to sea-land segmentation which is a difficult task. Due to general deep learning based methods required a large number of refined annotation, however, a large scale optical remote sensing images annotation is very expensive. Therefore, the Context-Unet is proposed to produce accurate sea-land segmentation results by a few annotated training samples. In this paper, we apply the Context-Unet network to the watershed extraction. On the basis of this, we re-compile the loss function to improve the accuracy of watershed extraction. Example: Finally, the date collected from Google Earth service is used to train and test this paper proposed Context-Unet and state-o-the-art methods. The experiments proved that the proposed method outperforms than other methods, and it can achieve 98% precision and 97% recall ratios.