基于局部背景对比的建筑外墙红外异常区域提取方法

刘政

doi:10.70693/cjst.v2i5.2175

Authors

刘政中国建筑标准设计研究院有限公司

DOI:

https://doi.org/10.70693/cjst.v2i5.2175

Keywords:

建筑外墙；红外热像；异常区域提取；局部背景；双阈值连接

Abstract

针对建筑外墙红外图像中日照、材料差异及构件边界形成的非均匀背景易导致全局分割方法误检和异常区域过度扩张的问题，提出一种基于局部背景对比与双阈值连接的红外异常区域提取方法。采用第2和第98百分位数对单幅图像进行稳健归一化，以大尺度邻域均值估计缓变背景，计算像素与局部背景的响应差；利用高、低阈值分别确定强响应核心和弱响应候选区域，保留与核心相连的8邻域连通域，并去除小面积孤立区域。该方法无需模型训练，参数在Building Facade RGB-IR Benchmark Dataset官方验证集上确定后保持固定。193幅测试图像上的精确率（Precision）、召回率（Recall）、F1值和交并比（IoU）分别为14.11%、46.51%、21.65%和12.14%。F1和IoU较K-means分别提高6.21和3.78个百分点，较Otsu分别提高8.21和4.94个百分点；无缺陷图像的平均异常面积占比由K-means的22.21%降至14.19%。结果表明，该方法可减少非均匀背景引起的大范围误检，使提取区域更加集中，但召回率低于两种对比方法，完全漏检率也较高，表明背景误检抑制与异常覆盖之间存在权衡。

References

[1] 彭雄, 钟新谷, 赵超, 等. 基于无人机热成像的建筑饰面层脱粘缺陷识别[J]. 红外技术, 2022, 44(2): 189-197.

[2] Tomita K, Chew M Y L. A review of infrared thermography for delamination detection on infrastructures and buildings[J]. Sensors, 2022, 22(2): 423.

[3] Balaras C A, Argiriou A A. Infrared thermography for building diagnostics[J]. Energy and Buildings, 2002, 34(2): 171-183.

[4] Grinzato E, Vavilov V, Kauppinen T. Quantitative infrared thermography in buildings[J]. Energy and Buildings, 1998, 29(1): 1-9.

[5] Barreira E, Freitas V P. Evaluation of building materials using infrared thermography[J]. Construction and Building Materials, 2007, 21(1): 218-224.

[6] Lucchi E. Applications of the infrared thermography in the energy audit of buildings: A review[J]. Renewable and Sustainable Energy Reviews, 2018, 82: 3077-3090.

[7] Garrido I, Barreira E, Almeida R M S F, et al. Introduction of active thermography and automatic defect segmentation in the thermographic inspection of specimens of ceramic tiling for building façades[J]. Infrared Physics & Technology, 2022, 121: 104012.

[8] Ibarra-Castanedo C, González D, Klein M, et al. Infrared image processing and data analysis[J]. Infrared Physics & Technology, 2004, 46(1-2): 75-83.

[9] Otsu N. A threshold selection method from gray-level histograms[J]. IEEE Transactions on Systems, Man, and Cybernetics, 1979, 9(1): 62-66.

[10] MacQueen J. Some methods for classification and analysis of multivariate observations[C]//Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability. Vol. 1. Berkeley: University of California Press, 1967: 281-297.

[11] Niblack W. An introduction to digital image processing[M]. Englewood Cliffs: Prentice-Hall, 1986.

[12] Sauvola J, Pietikäinen M. Adaptive document image binarization[J]. Pattern Recognition, 2000, 33(2): 225-236.

[13] He S, Zhao G, Chen J, et al. Weakly-aligned cross-modal learning framework for subsurface defect segmentation on building façades using UAVs[J]. Automation in Construction, 2025, 170: 105946.

[14] He S, Zhang S. Building Facade RGB-IR Benchmark Dataset[DS/OL]. DataSpace@HKUST, 2025[2026-09-20]. DOI: 10.14711/dataset/BEC8FI.