基于深度学习的混凝土试块表面纹络识别比对方法应用研究  

Research on the Application of Surface Texture Recognition Comparison Method of Concrete Test Block Based on Deep Learning

作　　者：何祖钧 黄俭 张兵 梁光世 黄杰 李伟 HE Zujun;HUANG Jian;ZHANG Bing;LIANG Guangshi;HUANG Jie;LI Wei(Foshan Sanshui District Construction Engineering Quality Inspection Station,Foshan Guangdong 528000,China;Guangzhou Yuejian Sanhe Software Co.,Ltd.,Guangzhou Guangdong 510000,China)

机构地区：[1]佛山市三水区建设工程质量检测站,广东佛山528000 [2]广州粤建三和软件股份有限公司,广东广州510000

出　　处：《工程质量》2024年第9期84-88,共5页Construction Quality

摘　　要：在混凝土试块整个周期中,试块的纹络具有唯一性,以此可以用来判断混凝土试块的同一性与真实性。论文对混凝土试块表面纹络进行了研究,并提出了一种基于深度学习的混凝土试块表面纹络识别比对模型,采用多分辨率识别、多尺度融合、注意力融合等方法,对不同时期的同一混凝土试块表面纹络进行识别和比对,得到试块表面纹络的差异度,以实现精确、高效的质量检验。Throughout the entire cycle of the concrete specimen,the texture of the specimen is unique,which can be used to determine the identity and authenticity of the concrete specimen.This article studies the surface texture of concrete specimen images and proposes a deep learning-based model for recognizing and comparing the surface texture of concrete specimen images.Using methods such as multi-resolution recognition,multi-scale fusion,and attention fusion,the surface texture of the same concrete specimen at different periods is recognized and compared to obtain the difference in surface texture of the specimen,in order to achieve accurate and efficient quality inspection.

关 键 词：混凝土试块 纹络 同一性 深度学习 多分辨率 

分 类 号：TU528[建筑科学—建筑技术科学]