Remote sensing image road extraction method and device

Abstract

The invention provides a remote sensing image road extraction method and device. The method comprises the following steps: firstly, acquiring a remote sensing image of a target area; then inputting the remote sensing image into a road extraction model, extracting and enhancing road contour features in the remote sensing image through the road extraction model, performing multi-scale feature extraction and fusion on the obtained road contour enhanced features, extracting road information of a target area based on the obtained multi-scale fusion features, and finally outputting the road information of the target area. By enhancing the road contour features and fusing the multi-scale features, the recognition capability of the road extraction model on the road contour can be enhanced, so that the integrity and accuracy of the road information obtained by recognition are better ensured.

Description

technical field [0001] The invention relates to the technical field of surveying and mapping, in particular to a method and device for extracting roads from remote sensing images. Background technique [0002] Road changes, including new road mileage, are very important basic data in road asset management and maintenance. At the same time, high-precision maps for unmanned driving also need to quickly update road information. Therefore, it has very important practical significance and research value to use new technology to quickly and objectively perceive the road status. [0003] Geographic information collection technologies, such as total stations and global positioning systems (Global Positioning System, GPS), are not only inefficient in collecting data on normal operating roads, but also can only obtain limited point data, which cannot fully display the whole picture of the road. Although Street View data provides clear road images, it is easily blocked due to the col...

AI Technical Summary

Problems solved by technology

[0004] Although remote sensing images can quickly and objectively obtain large-scale surface data, the road area accounts for a small proportion of the entire remote sensing image, and the shape of the road is variable, and the road material is different, which will lead to different colors and shapes of the road on the remote sensing image. In particular, roads extracted from remote sensing images often appear discontinuous and disconnected.

Images