基于深度学习的混凝土质量检测算法  

CONCRETE QUALITY DETECTION ALGORITHM BASED ON DEEP LEARNING

作　　者：刘辉宇[1] 杨震卿 黄爱菊 LIU Hui-yu;YANG Zhen-qing;HUANG Ai-ju(Huazhong University of Science and Technology,430074,Wuhan,China;Beijing Construction Engineering Group,100055,Beijing,China;BCEG Advanced Construction Materials Co.,Ltd.,102400,Beijing,China)

机构地区：[1]华中科技大学,武汉430074 [2]北京建工集团有限责任公司,北京100055 [3]北京建工新型建材有限责任公司,北京102400

出　　处：《建筑技术》2022年第7期957-961,共5页Architecture Technology

摘　　要：混凝土施工质量较大程度影响建筑物质量,为此开发了各种检测技术实现混凝土的质量检测。基于数字图像的特征提取方法以其高精度、高速度等特点得到广泛应用。基于深度学习的特征提取方法能够有效获取图像特征。基于此,利用摄像机拍摄清晰、完整的混凝土视频,选取兴趣帧,结合数字图像处理技术和深度学习的方法研究图像中混凝土的质量,通过综合考量坍落度和混凝土标号对混凝土质量的反映,提出了1种基于深度学习的混凝土质量检测算法。试验结果表明,该方法的混凝土质量检测准确率达72.23%。The quality of concrete construction greatly affects the quality of buildings,so various testing techniques are developed to realize the quality testing of concrete.Feature extraction based on digital image is widely used for its high precision and high speed.The feature extraction method based on deep learning can obtain image features effectively.Based on this,clear and complete concrete videos were shot with cameras,and interest frames were selected.The quality of concrete in images was studied by combining digital image processing technology and deep learning method.A concrete quality detection algorithm based on deep learning was proposed by considering the reflection of slump and concrete label on concrete quality.Experimental results showed that the accuracy rate of concrete quality detection could reach 72.23%.

关 键 词：深度学习 混凝土质量 检测方法 

分 类 号：TU753[建筑科学—建筑技术科学]