A method for coastline extraction from remote sensing images based on deep semantic segmentation network

Abstract

The invention discloses a remote sensing image coastline extraction method based on a deep semantic segmentation network, comprising the following steps: inputting coastal zone training remote sensing image samples into an encoder, extracting low-level detail features and high-level semantic features, and obtaining a multi-level feature map; The multi-level feature map is input to the decoder, and the image is decoded through the shoreline enhancement fusion module and the upsampling + convolution operation to obtain a set of sea and land segmentation feature maps and coastline distribution feature maps; pixel-by-pixel predictions are performed on the two sets of feature maps to obtain The binary map of sea and land segmentation and the binary map of coastline distribution are used to carry out error back-propagation using the binary labels of sea and land and coastline labels to obtain the trained network; the remote sensing images of the coastal zone are input into the trained network to obtain the segmentation of sea and land. Binary prediction image; extract the contour of the binary image as the sea-land boundary and remove the useless boundary, and then vectorize it to obtain the coastline. The invention can enhance the coastline distribution characteristics in the decoding process, and effectively improve the coastline extraction accuracy.

Description

technical field [0001] The invention relates to the field of remote sensing image information extraction, in particular to a remote sensing image coastline extraction method based on a deep semantic segmentation network. Background technique [0002] In recent years, with the rapid development of the marine economy and intensive human activities in coastal areas, the type and location of coastlines have changed significantly, coastal biodiversity has declined, and the ecological environment has been seriously affected. , Coastal change monitoring is of great significance. [0003] Coastline refers to the boundary between land and ocean. Due to the complex and diverse transitions between ocean and land, it is still a challenging problem to extract coastlines quickly and accurately from remote sensing images. At present, remote sensing image sea and land segmentation methods mainly include threshold segmentation method, edge detection operator method, object-oriented method, ...

AI Technical Summary

Problems solved by technology

Deep learning based on fully convolutional neural networks has achieved satisfactory performance in the field of semantic segmentation, but it is rarely applied to the coastline extraction task of remote sensing images

Images