扫地机器人U型墙+高反物品场景测试数据

扫地机器人在工作时时常会遇到一些特殊场景，比如U型墙场景且有高反物品存在的场景，基于视觉的扫地机在此场景中获取的深度数据常有边界不清晰、飞点多、多径干扰强等组合复杂问题，如何能在此场景中准确定位与避障是个难点也是重点，在此场景，扫地机深度视觉模组获取的深度数据的精度就显得十分重要。搭建扫地机真值系统，用真值相机和目标深度模组进行不同材质、光线、距离下U型墙+高反物品场景的数据采集，获取U形墙+高反物品场景的数据，并对数据进行处理得到校正后的数据用于目标深度模组的校正，提升深度模组在扫地机U型墙+高反物品场景下的测量精度，以解决扫地机定位与避障问题。采扫地机器人真值系统采集不同材质、光线、距离下的U型墙+高反物品场景的数据，包括原始深度相机数据、真值相机采集的点云数据；对原始深度相机数据进行系统误差去除；对真值相机采集的U型墙+高反物品场景点云数据进行修复，通过平面拟合算法进行U型墙+高反物品场景的墙面校正、再通过滤波算法对点云进行平滑去噪，得到修复后的真值相机点云；将修复后的真值相机点云与深度相机进行点云配准，得到校正后的U型墙+高反物品场景的点云数据——校正后数据laser-x、laser-y、laser-z和校正后数据tof-x、tof-y、tof-z。

Robotic vacuum cleaners frequently encounter special operational scenarios, particularly U-shaped wall scenarios containing high-reflectivity objects. Vision-based robotic vacuums often suffer from complex combined issues in the depth data collected in such scenarios, including unclear boundaries, excessive floating points, and severe multipath interference. Achieving accurate positioning and obstacle avoidance in these scenarios is both a key challenge and research priority, highlighting the critical importance of the accuracy of depth data acquired by the robotic vacuum's depth vision module. To address this, we constructed a robotic vacuum ground truth system, which uses a ground truth camera and the target depth module to collect data across U-shaped wall + high-reflectivity object scenarios under varying materials, lighting conditions, and distances. We obtained raw datasets of these scenarios, then processed the data to generate calibrated data for calibrating the target depth module, thereby improving the measurement accuracy of the depth module in U-shaped wall + high-reflectivity object scenarios for robotic vacuums, and solving the positioning and obstacle avoidance problems of robotic vacuums. This ground truth system collects data from U-shaped wall + high-reflectivity object scenarios under different materials, lighting conditions and distances, including raw depth camera data and point cloud data captured by the ground truth camera. First, systematic errors are removed from the raw depth camera data. Next, the point cloud data of the target scenarios collected by the ground truth camera is repaired: the wall surfaces in the U-shaped wall + high-reflectivity object scenarios are calibrated via plane fitting algorithms, followed by smoothing and denoising the point cloud through filtering algorithms, yielding the repaired ground truth camera point cloud. Finally, the repaired ground truth camera point cloud is registered with the depth camera's point cloud to obtain calibrated point cloud data for the U-shaped wall + high-reflectivity object scenarios, including the calibrated datasets laser-x, laser-y, laser-z and tof-x, tof-y, tof-z.

搜集汇总

数据集介绍

特点

该数据集包含扫地机器人在U型墙和高反物品场景下的测试数据，旨在提升机器人在复杂环境中的定位和避障能力。数据包括原始和校正后的激光和TOF数据，规模为1122条，每年更新一次。