High-resolution remote sensing image water body extraction method based on multi-scale optimization

Abstract

The invention discloses a high-resolution remote sensing image water body extraction method based on multi-scale optimization, and the method comprises the following steps: building a to-be-trained convolutional neural network, extracting multi-scale features from an input remote sensing image based on the network, and obtaining an initial rough water body segmentation result from the features with the lowest resolution; outputting a water body extraction result under full resolution through an attention erasing method in combination with the multi-scale features and the initial segmentation result; constructing a multi-scale loss function to obtain a trained convolutional neural network; and inputting a to-be-extracted high-resolution remote sensing image into the trained network to obtain a water body extraction result. According to the method, a remote sensing image training data set with real water body labels is learned and trained, and through the guidance of an attention erasingmechanism and in combination with a multi-scale optimization strategy, the identification and extraction of fine water bodies are enhanced while the overall water body extraction precision is remarkably improved.

Description

technical field [0001] The invention relates to the technical field of automatic extraction of remote sensing image information, in particular to a method for extracting water bodies from high-resolution remote sensing images based on multi-scale optimization. Background technique [0002] Water body extraction is a classic problem in the field of automatic extraction of remote sensing image information. Its main goal is to identify and extract water body regions in remote sensing images. The results of water body extraction are widely used in many fields, such as military reconnaissance, environmental protection, cartography and geographic analysis. Therefore, water extraction has important research value. [0003] Most of the traditional water body extraction methods are based on the water body index for water body extraction, which mainly uses the reflectance difference of the water body in different bands to identify the water body. However, these methods often misclas...

AI Technical Summary

Problems solved by technology

Due to the limitation of resolution, this method is usually difficult to capture the detailed information of the image, resulting in the inability to identify some small water bodies and inaccurate identification of edge areas.

Images