Remote Sensing Image Airport Runway Extraction Method Based on Joint Learning of Boundary and Context

Through the joint learning method of boundary and context, high-resolution network and adaptive optimization technology are used to solve the problem of insufficient runway boundary extraction in remote sensing images and the indistinguishable runway from the connecting track, achieving high-precision airport runway extraction.

CN116778315BActive Publication Date: 2025-07-04NORTHWESTERN POLYTECHNICAL UNIV
2 Cites 0 Cited by

Patent Information

Application Number
CN202211690241.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-12-28
Publication Date
2025-07-04
Estimated Expiration
2042-12-28

AI Technical Summary

Technical Problem

The prior art is difficult to finely extract airport runway boundaries and distinguish runways from connection channels in remote sensing images, especially in the absence of runway signs or typical characteristics.

Method used

Using a method based on boundary and context joint learning, multi-resolution features are extracted through high-resolution network HRNet, combined with boundary perception module and context aggregation module, the runway boundary and context features are generated, and the model is trained using adaptive moment estimation optimization and cross entropy loss function to realize runway segmentation.

Benefits of technology

The accuracy of runway boundary extraction and the ability to distinguish between runways and connecting tracks is improved. The generated runway segmentation map is almost completely aligned with the real boundary, which improves the model's scenario generalization ability.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present invention discloses a method for extracting airport runways from remote sensing images based on joint learning of boundaries and context, which solves the problems of inaccurate extraction results near the runway boundaries and difficulty in distinguishing runways from taxiways. The method mainly includes the following steps: obtaining an airport remote sensing dataset, annotating airport runway data, performing data augmentation on the annotated data to generate a training set and a test set; constructing a runway extraction network, using the High-Resolution Network (HRNet) as the feature extraction backbone to learn feature representations at four resolutions, inputting feature maps of different resolutions into a decoder, generating a runway boundary map through a boundary perception module, extracting context features through a context aggregation module, and finally outputting a runway extraction result, and training the runway extraction network using the training set; during the testing process, inputting the test data into the trained runway extraction network for forward prediction to obtain a runway segmentation prediction result.
Need to check novelty before this filing date? Find Prior Art