融合红外热成像与可见光图像的复合裂缝检测技术研究

打开文本图片集

中图分类号：TP391

文献标志码：A 文章编号：1001-5922（2025）12-0267-04

Research on composite crack detection technology combining infrared thermal imaging and visible light image

TANG Yan1，WANG Taol，SONGPengfeng²，YAO Zhongran （1.Wuxi Vocational and Technical University，Wuxi 214121，Jiangsu China; 2. Zhejiang Wanli University，Ningbo 3151Oo，Zhejiang China）

Abstract：Aiming at the problems of small-scale crack texture blurand edge artifacts in visible and infrared image fusion，this paper proposes a generative adversarial network（M2GAN）based on multi-scale feature extraction and multi-attention mechanism.The model constructs a multi-scale feature extraction module through the registered dual-modal input，and uses the side-to-side connection fusion strategy to simultaneously retainthe texture details of thevisible lightimageandthethermalradiationsemantic informationof theinfrared image to enhancethecharacterization abilityof thecrack area.Experiments onthebuilding surface defect dataset showthat M2GAN is significantly better than the mainstream methods in terms of structural similarityand edge retention，with an average increase of 10.66% and 24.92% ，which effectively solves the problem of crack detail loss in weak light environment and comprehensively enhances the diagnostic accuracy of composite images in structural health monitoring.

Key Words：infrared thermography;visible light image;composite crack detection technology;multi scale feature extraction

随着交通基础设施使用年限不断增长，路面裂缝检测对于保障道路安全具有至关重要的作用，传统单模态检测方法受到光照、天气等环境因素，难以在复杂工况下稳定提取裂缝特征。（剩余6070字）