路面服役性能全息感知装备研发与高通量数据挖掘技术数据集

沥青路面典型病害与薄层罩面连续磨耗的高密高精三维激光扫描数据与指标1.3 “a）路表病害的自动识别较传统二维检测人工识别提高4倍以上，误识别率低于10%”“以及指标10.2实现路表纹理垂直精度不大于0.5mm，点云数据横纵间隔不大于1.0mm”相对应，是课题五进行科学实验的核心数据，支撑了学术论文、研究生毕业论文、以及专利等。
数据集的内容：本数据集整合了多类型数据，为路面损伤特征研究与服役性能分析提供了丰富资料。多源原始数据采集主要包括三维深度图像、灰度图片以及雷达频率数据。利用车载式三维相机拍摄获取图像数据，包含横向裂缝、网裂、车辙、坑槽、松散、拥抱、剥落等典型病害。同时，通过三维探地雷达采集雷达频率数据，涵盖路面内部结构信息，包括不同频率下的数据，以检测内部病害和分层结构。内外部损伤数据处理主要包括三维点云数据和雷达处理后的波纹图。对采集到的点云数据进行滤波、去噪和配准处理，剔除异常点，确保数据完整性和精确性，形成可用于分析的高精度三维点云数据，反映路面表面和内部结构特征。多维度指标数据构建包括三维车辙各类指标数据、纹理参数等。基于小波矩和目标矩理论，提取路面表面纹理参数，如纹理深度、纹理方向等，反映路面表面特性。通过采用自适应方法生成匹配模板，实现了多层位、多尺度、多特征的路面服役性能全要素动态挖掘。
数据类型：xlsx，csv，jpg，time，cor，iprb
采集方案：对于现场三维点云数据采集，通过数据处理算法（如滤波、配准等），消除噪声并优化数据质量，确保采集到的点云和图像保持高精度且具有较好的数据对齐效果。通过对比、截断与数据插补，最终形成一个包含丰富病害特征、完整性强且噪声少的三维数据集。通过去噪、优化及网格化处理，消除激光扫描过程中产生的噪声点，确保点云数据的完整性和精确性。针对采集到的数据进行对齐和配准，确保不同磨耗阶段和抛丸处理前后的表面特征可以精确对比分析。数据的进一步优化包括对少量缺失点的插补处理，以确保三维模型的完整。
采集地点：内蒙古呼和浩特市G6卧佛山隧道、G5901微表处、G110国道、G301国道、陕西西安市长安大学道路所315
采集时间：2023年1月-2024年4月
设备情况：车载式三维激光扫描设备、室内三维激光扫描系统、高算力工作站，高性能工作站，探地雷达检测车。

High-density and high-precision 3D laser scanning data and indicators for typical asphalt pavement diseases and continuous wear of thin-layer overlays 1.3: "a) The automatic identification of pavement surface diseases is more than 4 times faster than manual identification via traditional 2D detection, with a false recognition rate lower than 10%"; and indicator 10.2 realizes that the vertical accuracy of pavement surface texture is no more than 0.5 mm, and the horizontal and vertical spacing of point cloud data is no more than 1.0 mm. This is the core data for scientific experiments conducted in Project 5, which supports academic papers, graduate theses, patents, etc.

This dataset integrates multi-type data, providing rich resources for the research of pavement damage characteristics and service performance analysis. Multi-source raw data collection mainly includes 3D depth images, grayscale images, and radar frequency data. Image data are captured by vehicle-mounted 3D cameras, covering typical pavement diseases such as transverse cracks, alligator cracks, rutting, potholes, raveling, upheaval, and spalling. Meanwhile, radar frequency data are collected via 3D ground-penetrating radar (GPR), covering pavement internal structural information including data under different frequencies for detecting internal diseases and layered structures. Internal and external damage data processing mainly includes 3D point cloud data and radar-processed ripple maps. The collected point cloud data are processed through filtering, denoising, and registration to remove outliers, ensuring data integrity and accuracy, thus forming high-precision 3D point cloud data available for analysis, which reflects the surface and internal structural characteristics of the pavement. Construction of multi-dimensional indicator data includes various 3D rutting indicator data, texture parameters, etc. Based on the theories of wavelet moments and target moments, pavement surface texture parameters such as texture depth and texture direction are extracted to reflect pavement surface properties. By using an adaptive method to generate matching templates, dynamic mining of all elements of pavement service performance with multiple layers, multiple scales, and multiple features is realized.

Data types: xlsx, csv, jpg, time, cor, iprb

For on-site 3D point cloud data collection, data processing algorithms (such as filtering, registration, etc.) are used to eliminate noise and optimize data quality, ensuring that the collected point clouds and images maintain high precision and good data alignment effects. Through comparison, truncation, and data interpolation, a 3D dataset with rich disease features, strong integrity, and low noise is finally formed. Through denoising, optimization, and gridding processing, noise points generated during laser scanning are eliminated to ensure the integrity and accuracy of point cloud data. The collected data are aligned and registered to ensure accurate comparative analysis of surface features at different wear stages and before and after shot blasting treatment. Further optimization of the data includes interpolation processing for a small number of missing points to ensure the integrity of the 3D model.

Collection locations: G6 Wolufoshan Tunnel, G5901 Micro-Surfacing Section, G110 National Highway, G301 National Highway in Hohhot, Inner Mongolia Autonomous Region; and Road Institute 315 of Chang'an University, Xi'an, Shaanxi Province

Collection period: January 2023 to April 2024

Equipment: Vehicle-mounted 3D laser scanning equipment, indoor 3D laser scanning system, high-performance computing workstations, and ground-penetrating radar detection vehicle.

搜集汇总

背景与挑战

背景概述

该数据集整合了多源原始数据，包括三维深度图像、灰度图片和雷达频率数据，用于路面损伤特征研究与服役性能分析。数据采集于2023年1月至2024年4月，地点覆盖内蒙古呼和浩特市及陕西西安市的多条道路，通过车载式三维相机和探地雷达设备获取。数据集包含多种文件格式，总数据量为1.98GB，共7851个文件，支撑了沥青路面典型病害的自动识别与纹理精度分析。