基于红外图像特征的建筑外墙热工缺陷快速识别方法  

A Rapid Identification Method for Thermal Defects of Building Exterior Walls Based on Infrared Image Features

作　　者：鲁友存 端木琳[1] 王宗山[1] LU Youcun;DUANMU Lin;WANG Zongshan(School of Civil Engineering,Dalian University of Technology,Dalian 116024,Liaoning,China)

机构地区：[1]大连理工大学土木工程学院,辽宁大连116024

出　　处：《建筑科学》2023年第6期1-9,共9页Building Science

基　　金：“十三五”国家重点研发计划项目“绿色宜居村镇技术创新”(2018YFD1100705)。

摘　　要：既有建筑外墙热工缺陷定期普查工作对减少公共安全隐患和降低建筑能耗有非常重要的影响。目前红外热像法在建筑外墙热工缺陷检测中得到广泛应用,但其在后期图像判断过程中存在主观性强、工作量大、效率低下的问题。为了克服以上问题,本文提出了1种基于红外图像特征的建筑外墙热工缺陷快速识别方法。该方法在选取并分析正常墙体和热工缺陷墙体红外图像灰度直方图统计特征、方向梯度直方图(HOG)特征和局部二值模式(LBP)特征3类具有强可分性特征的基础上,利用反向传播(BP)神经网络和支持向量机(SVM)分类器对输入的图像进行训练测试。根据测试得到的分类准确率确定了局部二值模式特征和支持向量机分类器结合的分类模型,作为墙体正常与否的判断方式。相较于检测规程中的方法,该方法消除了主观因素的影响,能够实现批量快速地处理图像,热工缺陷识别的准确率和效率得到大幅提高,为既有建筑外墙热工缺陷定期普查工作提供了新的技术思路。The regular inspection for thermal defects of existing building exterior walls has a major impact on reducing public safety hazards and lowering building energy consumption.At present,infrared thermal imaging method has been widely used in the thermal defect detection of building exterior walls,but it faces problems of strong subjectivity,heavy workload,and low efficiency in the later image judgment process.In order to overcome the above problems,a rapid identification method for thermal defects of building exterior walls based on infrared image features was proposed in this paper.Based on the selection and analysis of gray histogram statistical features,directional gradient histogram(HOG)features and local binary pattern(LBP)features of infrared images of normal walls and thermal defect walls,this method used back propagation(BP)neural network and support vector machine(SVM)classifier to train and test the input images.According to the classification accuracy obtained from the test,the classification model combining local binary pattern features and support vector machine classifier was determined as the judgment method of whether the wall is normal or not.Compared with the method in the detection specification,this method eliminates the influence of subjective factors,and can realize batch and rapid image processing,significantly improving the accuracy and efficiency of thermal defect detection,and providing a new technical idea for the periodic inspection of thermal defects of existing building exterior walls.

关 键 词：热工缺陷 红外图像 图像特征 局部二值模式 支持向量机 

分 类 号：TU111.4[建筑科学—建筑理论] TU17