长沙理工大学学报（自然科学版）2025，Vol.22Issue(6)：46-58,13.DOI:10.19951/j.cnki.1672-9331.20250814001
基于电场强度稀疏深度网络的结构裂缝三维检测方法
Three-dimensional detection of structural cracks based on sparse deep networks of electric field intensity
摘要
Abstract
[Purposes]Non-destructive detection of structural cracks in expressway pavement is a crucial link in supporting road maintenance and traffic operation safety.Although detection methods based on ground-penetrating radar(GPR)images and deep neural networks have achieved certain results,there are still deficiencies in terms of accuracy and stability.Therefore,this paper optimized the intelligent detection performance of GPR and focused on addressing the issues of accuracy and stability in detecting structural cracks in pavement,providing technical support for road maintenance.[Methods]A three-dimensional detection method for structural cracks in pavement based on the sparse deep network of electric field intensity was proposed,namely sparse convolutional GPR detection network.Firstly,this method directly took the three-dimensional electric field intensity distribution of GPR as input and converted it into a three-dimensional voxel tensor.Then,the three-dimensional voxel tensor was successively processed by the SparseNorm module,Dual module,and Tri module to extract the structural crack features in pavement from the electric field intensity distribution.Finally,the structural crack features in pavement were input into the separated detection head to predict the three-dimensional bounding box of the structural crack in pavement.[Findings]The experimental results on three expressways show that the average precision rates of the detection method at thresholds of 0.3 and 0.5 are 0.831 and 0.463,respectively,and its accuracy and stability are both better than those of the image detection network.[Conclusions]The sparse deep neural network with the electric field intensity distribution of GPR as input provides a high-precision automatic method for non-destructive detection of structural cracks in expressway pavement.关键词
路面检测/雷达信号检测处理/裂缝检测/探地雷达/深度神经网络Key words
pavement detection/radar signal detection and processing/crack detection/ground-penetrating radar/deep neural network分类
交通工程引用本文复制引用
侯兆军,方亮文,黄炜,童峥..基于电场强度稀疏深度网络的结构裂缝三维检测方法[J].长沙理工大学学报（自然科学版）,2025,22(6):46-58,13.基金项目
国家自然科学基金项目(52308447) （52308447）
苏州市市政管理中心技术开发项目(SZSZ-2024-1-KJ) （SZSZ-2024-1-KJ）
江苏省青年科技人才托举工程项目(JSTJ-2024-089) National Natural Science Foundation of China(52308447) （JSTJ-2024-089）
Suzhou Municipal Administration Center(SZSZ-2024-1-KJ) （SZSZ-2024-1-KJ）
Young Scientific and Technological Talents Promotion Programme of Jiangsu Province(JSTJ-2024-089) （JSTJ-2024-089）
