摘要
脱空缺陷在水泥路面中不可避免,严重影响水泥路面的结构安全性,迫切需要建立脱空缺陷早期识别定位方法,为路面精准预养护提供依据.根据探地雷达(GPR)数据的成像原理,提出了标准化和去背景的脱空特征增强算法.依据GPR发射Ricker子波的工作特点,借鉴深度学习中卷积网络的特征提取特性,以A-Scan数据集为对象,设计了提取脱空特征的二维卷积核,并提出基于卷积核和阈值判断的脱空定位算法.采用数值模拟和室内验证相结合的方法,研究了不同卷积尺度的影响,并将设计的卷积识别算法在水泥路面进行现场验证.实验结果表明,提出的标准化和去背景的预处理算法可替换传统的GPR多步骤数据处理的算法组合,有效实现脱空区域的特征增强作用;卷积算法有效突出了脱空特征,卷积核尺度过大会降低图谱分辨率,卷积尺度过小无法突出脱空特征,合适卷积纵向尺寸为1~1.5倍f_(s)/f_(0),横向尺寸根据检测对象一般取对应实际距离10~20 cm的采样点.实验统计方法设置了卷积核判断阈值,可有效确定脱空区域的深度以及判断脱空区域是否含水,为水泥路面病害的自动识别奠定基础.
Void defects are inevitable in cement pavement,which seriously affects the structural safety of cement pavement.There is an urgent need to establish an early recognition method for void defects to provide a basis for accurate road maintenance.According to the imaging principle of Ground Penetrating Radar(GPR)data,a standardization and background subtraction feature enhancement algorithm is proposed.Considering the working characteristics of GPR sending Ricker wavelets,drawing on the feature extraction characteristics of convolutional networks in deep learning,taking the A-Scan data set as the object,designing a two-dimensional convolution kernel for extracting void features,and proposing based on convolution Void positioning algorithm based on core and threshold judgment.Experimental results show that the proposed preprocessing algorithm for standardization and background subtraction can replace the traditional GPR multi-step data processing algorithm combination,and effectively realize the feature enhancement effect of the void area;The convolution algorithm effectively highlights the void features.If the convolution kernel scale is too large,it will reduce the resolution of the map.If the convolution scale is too small,the void features cannot be highlighted.The appropriate vertical size of the convolution is 1~1.5 times f_(s)/f_(0),and the horizontal size is based on the detection.Objects generally take sampling points corresponding to the actual distance of 10~20 cm.The experimental statistical method sets the convolution kernel judgment threshold,which can effectively determine the depth of the void area and determine whether the void area contains water,which lays the foundation for the automatic identification of cement pavement diseases.
作者
罗婷倚
杨哲
张军
余秋琴
朱欣
LUO TingYi;YANG Zhe;ZHANG Jun;YU QiuQin;ZHU Xin(Gunagxi Beitou Highway Construction and Investment Group Co.,Ltd.,Nanning 530028,China;National Engineering Laboratory for Highway Maintenance Equipment,Xi'an 710064,China;Key Laboratory for Highway Construction Technology and Equipment of Ministry of Education,Xi'an 710064,China)
出处
《地球物理学进展》
CSCD
北大核心
2022年第6期2580-2588,共9页
Progress in Geophysics
基金
广西交通运输行业重点科技项目(19-09)
陕西省2022年自然科学基础研究计划(2022JM-249)
陕西省交通厅项目(20-30X)联合资助。
关键词
道路工程
水泥路面
脱空检测
探地雷达
特征增强
Road engineering
Cement pavement
Void detection
Ground penetrating radar
Feature enhancement
