An edge attention enhanced double-branch framework building contour extraction method and device

By using a dual-branch frame field network with edge attention enhancement, the accuracy and efficiency issues of building contour extraction in complex scenes are solved, achieving high-precision and structurally consistent building vectorization extraction, which is suitable for remote sensing mapping and urban modeling.

CN120894566BActive Publication Date: 2026-03-10AEROSPACE INFORMATION RES INST CAS
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-07-15
Publication Date
2026-03-10

AI Technical Summary

Technical Problem

Existing building contour extraction methods lack accuracy and computational efficiency in complex scenes, making it difficult to meet real-time processing requirements. Furthermore, insufficient boundary feature extraction can easily lead to contour breakage or distortion.

Method used

We employ an edge attention-enhanced dual-branch frame field network (EA-DBFFN) to generate structurally accurate building vector maps through multi-scale feature extraction, dual-task parallel prediction, mask and edge feature interaction fusion, frame field decoding, and path optimization.

Benefits of technology

It significantly improves the accuracy and structural integrity of building outline extraction, increases computational efficiency, enhances robustness in complex scenarios, and realizes end-to-end automated processing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120894566B_ABST
    Figure CN120894566B_ABST
Patent Text Reader

Abstract

The application discloses a kind of edge attention enhanced double-branch framework field building outline extraction method and device, belong to computer vision and remote sensing image processing technical field.The method adopts edge perception double-branch framework field network (EA-DBFFN), and building outline extraction of end to end is realized by image input, feature extraction, double-task decoding, mask framework field fusion and polygonization.Wherein, double-task decoding includes mask branch and edge branch, respectively through dynamic receptive field module and multi-directional convolution module to enhance feature expression ability, and using double-task mutual guidance realizes the feature interaction and collaborative optimization between two branches.The application strengthens boundary feature learning by edge attention mechanism, and significantly improves the structural integrity and boundary precision of building outline extraction by combining the geometric constraint of framework field vector field.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Image processing method and device, electronic equipment and computer readable medium

    CN113112422A

  • High-resolution remote sensing image building vector contour data extraction method

    CN116682021A