Robot Quadruped Material Database

We present the RQMD dataset, a comprehensive collection of diverse material samples aimed at advancing computer vision and machine learning algorithms in terrain classification tasks. This dataset contains RGB images of 5 different terrains, such as Asphalt, Brick, Grass, Gravel, and Tiles, captured using an 8-megapixel Raspberry Pi camera from a top-view perspective. Notably, the dataset encompasses images taken at different times of the day, introducing variations in lighting conditions and environmental factors.To facilitate friction force extraction, we conducted an experiment using a rubber ball, simulating the foot of a small-scale robot. The experiment involved scratching the rubber ball on the different terrains using a DC motor with an encoder, allowing us to measure the friction force acting on the ball. This information contributes to a more comprehensive understanding of the terrain properties.In summary, RQMD serves as a valuable resource for researchers and practitioners seeking to enhance their computer vision and machine learning models for terrain classification tasks, and the friction force extraction experiment adds an additional dimension to the dataset, enabling deeper insights into terrain characteristics.

我们提出RQMD数据集，这是一套涵盖多样材质样本的综合性数据集，旨在推进地形分类任务中的计算机视觉与机器学习算法研究。该数据集包含5类不同地形的RGB（红-绿-蓝）图像，分别为沥青（Asphalt）、砖块（Brick）、草地（Grass）、碎石（Gravel）与瓷砖（Tiles），由800万像素树莓派（Raspberry Pi）相机以俯视视角采集得到。值得注意的是，该数据集涵盖了一日内不同时段拍摄的图像，引入了光照条件与环境因素的多样性变化。为支撑摩擦力提取相关研究，我们开展了一项实验：以橡胶球模拟小型机器人的足部。实验通过搭载编码器（encoder）的直流电机（DC motor）驱动橡胶球在各类地形表面刮擦，以此测量作用于球体的摩擦力。该信息有助于更全面地认知地形属性。综上，RQMD数据集可为致力于优化地形分类任务相关计算机视觉与机器学习模型的研究者与从业者提供宝贵的资源支持，同时摩擦力提取实验为该数据集增添了额外维度，助力研究者更深入地剖析地形特征。

搜集汇总

背景与挑战

背景概述

Robot Quadruped Material Database是一个包含五种地形RGB图像的数据集，用于计算机视觉和机器学习的地形分类任务。数据集还包括通过摩擦实验测量的摩擦力数据，为地形特性研究提供了额外维度。