Laser-scanned roadway range image dataset (LRRD)

This experiment aims at generating a laser-scanned range (i.e., elevation) image dataset for roadway crack classification. A vehicle-mounted laser imaging system called Laser Image and Measurement System was utilized for roadway image data collection. On 12/30/2019, the experiment was performed on asphalt and concrete roadways in Northport, Alabama. A dataset was obtained, which consists of 5000 crack image patches (dimension: 256-by-256 pixels) and 5000 non-crack patches cropped from the acquired roadway image frames. This range image dataset can be used for training and testing DCNN classifiers.

本实验旨在构建用于道路裂缝分类的激光扫描距离（即高程）图像数据集。本研究采用一款名为激光成像与测量系统（Laser Image and Measurement System）的车载激光成像系统开展道路图像数据采集工作。本次实验于2019年12月30日在美国阿拉巴马州北港地区的沥青及混凝土道路上完成。最终获取的数据集包含5000张裂缝图像块（尺寸：256×256像素）与5000张非裂缝图像块，两类图像块均从采集得到的道路图像帧中裁剪获得。该距离图像数据集可用于深度卷积神经网络（Deep Convolutional Neural Network，DCNN）分类器的训练与测试。

搜集汇总

背景与挑战

背景概述

LRRD数据集包含10000个激光扫描的道路范围图像块（5000裂缝和5000非裂缝），用于DCNN道路裂缝检测研究，数据采集自沥青和混凝土路面，图像尺寸为256x256像素。