结合纹理特征的深度学习混凝土裂缝识别方法  

Deep learning concrete crack identification method combined with texture features

作　　者：潘远 周双喜[1] 黄晓生[1] 杨丹[1] 丁杨[2,3] PAN Yuan;ZHOU Shuangxi;HUANG Xiaosheng;YANG Dan;DING Yang(East China Jiaotong University,Nanchang 330013,China;Zhejiang University,Hangzhou 310058,China;Zhejiang University City College,Hangzhou 310015,China)

机构地区：[1]华东交通大学,江西南昌330013 [2]浙江大学,浙江杭州310058 [3]浙大城市学院,浙江杭州310015

出　　处：《混凝土》2024年第2期45-51,57,共8页Concrete

基　　金：国家自然科学基金(51968022);省主要学科学术和技术带头人(20213BCJL22039)。

摘　　要：混凝土的结构裂缝对结构安全性会产生重要影响,因而对混凝土裂缝进行检测监测是非常必要的。针对深度学习方法需要大量的训练样本,深度学习模型训练过程需要耗费大量时间的特性,提出一种将纹理特征与混凝土裂缝数据融合的深度学习目标检测框架。将纹理特征与预处理混凝土数据进行多数据融合,增加训练样本的特征通道数,减少模型对训练样本需求,提高模型训练速度;并在此基础上实现样本数量有限情况下的混凝土裂缝检测。通过在自制钢纤维混凝土裂缝数据集上进行试验,将该目标检测框架与未进行纹理特征融合以及与未进行预处理混凝土数据的目标检测方法进行对比。试验结果表明:本研究提出的裂缝检测方法,不仅可以大大减少训练模型所需时间,在模型训练样本较少时亦可获得较好的检测效果,检测精度也得到一定的改善。The structural cracks of concrete have a significant influence on the structural safety,so it is necessary to detect and monitor concrete crack.Aiming at masses of training samples are needed for deep learning methods,deep learning model also need large time to train model,a deep learning target detection framework combining texture features with concrete crack data is proposed.Texture features and pre-processed concrete data are merged to increase the number of feature channels in order to reduce the demand of training samples for the model and improve training speed.Concrete crack detection can be realized under the condition of limited number of samples.Through self-made steel fiber reinforced concrete crack data set to experiment and comparison.The experimental results show that the number of parameters that need to be fitted in the model training and training time can be correspondingly reduced and the detection accuracy has also been improved.

关 键 词：纹理特征 深度学习 裂缝识别 

分 类 号：TU528.01[建筑科学—建筑技术科学]