A geographic information based sea-land segmentation method for HR optical remote sensing image

In this letter, a novel method is proposed to conduct sea-land segmentation in the help of geographic information for high resolution (HR) optical remote sensing image. Firstly the pre-segmentation is conducted by the means of image analysis. Then the salient structure mainly formed by artificial facility can be used to determine the region of interests (ROI), which contains the potential ship targets. After that, accurate segmentation is made in ROI via mask generated by geographic information and structure feature. Experiments on Google image shows that the method proposed in the paper performs better than existing methods without the geographic information, particularly for the berth area where ship targets gather.