基于YOLOv4算法的建筑大面积现浇混凝土地坪施工细小裂缝检测  

Detection of small cracks in large area cast in place concrete flooring construction based on YOLOv4 algorithm

作　　者：卢亚荣 LU Yarong(Shaanxi Construction Science Research Institute Co.,Ltd.,Xi'an 710082,China)

机构地区：[1]陕西省建筑科学研究院有限公司,西安710082

出　　处：《无损检测》2025年第4期39-43,共5页Nondestructive Testing

摘　　要：受混凝土材料收缩以及温度等作用效应,建筑施工中混凝土地坪结构开裂问题较为普遍,会直接影响到地坪底部地基的稳定性。传统方法对裂缝结构特征的提取方式较为单一,在对地坪施工细小裂缝检测中存在误差,为此,基于YOLOv4算法研究建筑大面积现浇混凝土地坪施工细小裂缝检测方法。首先在温度场下定量分析地坪施工形变分量;其次确定大面积地坪形变分量中的裂缝信息;再采用均值偏移滤波聚类细小裂缝特征;最后基于YOLOv4算法融合细小特征检测混凝土地坪细小裂缝。试验结果表明,所研究方法可以较为完整地提取混凝土地坪细小裂缝特征,在不同类型裂缝目标检测中精度可达98%,具有应用价值。Due to the shrinkage of concrete materials and temperature effects,cracking of concrete floor structures is common in construction,which directly affects the stability of the foundation at the bottom of the floor.The traditional method for extracting crack structure features is relatively single,and there are errors in detecting small cracks in floor construction.Therefore,based on YOLOv4,a method for detecting small cracks in large-area cast-in-place concrete floor construction was studied.The deformation component of floor construction under temperature field was quantified,the crack information in the deformation component of large-area flooring was determined.Mean shift filtering was used to cluster small crack features,small cracks in concrete floors were detected based on YOLOv4 fusion of small features.The results showed that the proposed method could extract the characteristics of small cracks in concrete floors more comprehensively,with an accuracy of up to 98%in target detection of different types of cracks,and had practical value.

关 键 词：YOLOv4算法 混凝土地坪 细小裂缝 

分 类 号：TU755.7[建筑科学—建筑技术科学] TG115.28[金属学及工艺—物理冶金]