A building damage classification method based on post-disaster unmanned aerial vehicle remote sensing image

By combining the feature encoding and classification modules of a convolutional neural network with spatial attention and global context feature extraction, the problem of insufficient accuracy in post-disaster building damage assessment is solved, and high-accuracy damage level classification is achieved.

CN115331049BActive Publication Date: 2026-03-31SHANGHAI OCEAN UNIV
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-08-01
Publication Date
2026-03-31

AI Technical Summary

Technical Problem

Existing post-disaster building damage assessment methods suffer from insufficient accuracy in single-phase assessments, particularly in distinguishing between minor and severe damage. Furthermore, the practicality of existing dual-phase methods is limited by the impact of image collection.

Method used

A convolutional neural network for building damage classification (EBDC-Net) is adopted, which combines a feature encoding module and a classification module. It utilizes a spatial attention mechanism to gather similar features and combines global and contextual feature extraction to improve the accuracy of damage level classification.

Benefits of technology

It achieves accurate classification of post-disaster building damage levels, with classification accuracy rates of 94.44%, 85.53%, and 77.49%, respectively, classifying buildings with different degrees of damage to meet the needs of rapid and accurate assessment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115331049B_ABST
    Figure CN115331049B_ABST
Patent Text Reader

Abstract

The application discloses a building damage classification method based on post-disaster unmanned aerial vehicle remote sensing images, acquires post-disaster unmanned aerial vehicle remote sensing images, and carries out pretreatment; a building damage classification network model is established and trained; the network model comprises a feature coding module and a classification module, the feature coding module comprises a plurality of convolution blocks connected in sequence and provided with an attention mechanism, is used for carrying out feature coding on the unmanned aerial vehicle remote sensing images, acquires a feature coding graph, the classification module comprises a global feature extraction module, a context feature extraction module and a classifier, the global feature extraction module is used for carrying out global feature extraction on the feature coding graph, the context feature extraction module is used for carrying out context feature extraction on the feature coding graph, and the classifier is used for classifying image features fused with global features and context features; the trained building damage classification network model is used for damage classification and identification of the acquired post-disaster unmanned aerial vehicle remote sensing images.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Remote-sensing image building change detection method

    CN110705457A