A Deep Learning Method for Simultaneously Extracting Road Surface and Centerline from Remote Sensing Images
A technology of road pavement and remote sensing images, applied in neural learning methods, character and pattern recognition, instruments, etc., can solve problems such as incomplete roads and poor road connectivity, so as to improve the degree of automation, enhance connectivity, and reduce misjudgments Effect
Active Publication Date: 2022-02-01
WUHAN UNIV
View PDF9 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0003] Aiming at the deficiencies of the prior art, the present invention provides a road network extraction framework for remote sensing images based on convolutional neural networks, which can simultaneously extract road surfaces and centerlines in remote sensing images, fully combining image semantic segmentation methods and centerline tracking methods , achieve complementary advantages, overcome the shortcomings of poor road connectivity in road segmentation results and incomplete roads in single-point tracking results, and obtain high-quality road network data with good integrity and topological connectivity
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0036] The specific implementation manner of the present invention is illustrated below through examples and accompanying drawings.
[0037] An embodiment of the present invention provides a deep learning method for simultaneously extracting road surfaces and centerlines from remote sensing images, including the following steps:
[0038] Step 1, build a sample library based on existing images and corresponding annotation files, including road segmentation dataset and road centerline tracking dataset; Step 2, use the road segmentation dataset constructed in step 1 to construct the road segmentation network D-LinkNet Carry out training, and then perform pixel-level prediction on the road in the remote sensing image to obtain the initial segmentation result of the road surface, and on this basis, combine the promotion strategy in machine learning to design a network model for improving the segmentation result (Boosting Segmentation Network, BSNet), train a number of improved segm...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention relates to a new deep learning method for simultaneously extracting road surfaces and center lines of remote sensing images. Give full play to the advantages of convolutional neural network in road surface segmentation and road centerline tracking, and use the road surface and centerline results to constrain each other and complement each other to extract road networks from remote sensing images, which solves the problem of incomplete road results in previous extraction methods. Poor connectivity etc. The invention has the following advantages: strong robustness, adaptable to road extraction from remote sensing images of different scales, sustainable iteration and continuous optimization, not only can extract road surfaces with topological connectivity, but also can extract accurate and complete road centerlines, and can be applied In the fields of urban planning, driving navigation, disaster emergency response, map production and update, etc.
Description
technical field [0001] The invention relates to a method for automatically extracting roads from remote sensing images based on convolutional neural networks, which can simultaneously extract road surfaces and centerlines, effectively improve the integrity and connectivity of road extraction, and assist in the construction and update of road networks. Data can be widely used in urban planning, autonomous driving and other fields. Background technique [0002] As a kind of basic geographic data, road network plays an important role in driving navigation, disaster emergency response, map drawing and updating, etc. However, the construction and update of road network data still rely on manual operations, which is time-consuming and laborious. As one of the most important earth observation technologies for obtaining geometric and physical information of the earth's surface, remote sensing technology has been developed rapidly, which has drawn more and more attention to extracti...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06V20/10G06V10/774G06V10/764G06K9/62G06N3/04G06N3/08G06V10/82
CPCG06N3/08G06V20/194G06V20/182G06N3/045G06F18/241G06F18/214
Inventor 季顺平魏瑶张凯
Owner WUHAN UNIV