项目名称: 基于激光扫描数据的路面裂缝连续检测方法研究
项目编号: No.51208392
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 建筑环境与结构工程学科
项目作者: 黄玉春
作者单位: 武汉大学
项目金额: 25万元
中文摘要: 高精度激光扫描路面数据具有直接、分辨率高、抗干扰能力强、对光照不敏感、可高速连续采集等特性,它的出现为研究路面裂缝检测带来了新的机遇。本项目以高精度激光扫描路面数据为基础,提出局部路面裂缝的裂纹几何元表示及其运算规则, 实现快速、准确、连续增量式检测路面裂缝。创新之处在于: 1) 基于逐条扫描线的小波逼近和空间双边滤波的激光扫描数据预处理,可有效去除噪声及缓慢变化的车辙等影响,增加路面裂缝的对比度;2) 从局部裂缝几何角度出发,提出基于Hessian二阶结构分析的裂纹几何元表示与裂缝增强,无需任何二值化分割处理,可全面提取局部路面裂缝的中心、方向、宽度、长度和深度等几何信息,鲁棒性强且易于并行方式快速处理;3) 利用混合高斯模型可有效实现裂纹几何元的并和交运算,消除局部裂纹几何元的分块间隙,保证路面裂缝的连续式检测。本项目研究成果可供路面同步养护处理(如封缝)参考。
中文关键词: 道路裂缝;激光扫描;裂缝几何元;张量投票;Hessian结构分析
英文摘要: High-accuracy laser profiling data of the pavement are direct, of high-resolution, robust to interference, insensitive to illuminatin,able to be collected fast and continuously, and bring new opportunity to the study of pavement crack. Based on the high-accuracy pavement profiling data, this project proposes Geometric Crack Element (GCE) and its calculus to characterize the local pavement crack and to detect pavement crack in a fast, accurate,and continuous way. The innovations of this project include: 1) preprocessing of laser profiling data through profile-by-profile Wavelet approximation and bilateral filtering, which can effectively reduce noise and low-frequency rutting, and enhances the contrast of pavement crack; 2)GCE and crack enhancement through Hessian second-order structure, which is based on local crack geometry, and able to robustly characterize the center, orientation, width, length, and depth of the local crack in a parallel way; 3) Gaussian Mixture Model (GMM) -based GCE calculus, which can eliminate the analysis gap between GCEs and ensures continuous detection of pavement crack. The outcome of this project can guide the synchronous pavement maintenance (e.g., crack sealing).
英文关键词: Pavement Crack;Laser Profiling;Geometric Crack Element (GCE);Tensor Voting;Hessian Structure Analysis