室外场景视觉定位数据

面向基于视觉的机器人环境建模与定位导航基础研究，为突破复杂环境下机器人视觉实时高精度定位技术瓶颈，基于移动视觉机器人平台的双目RGB彩色相机，在同济大学校园内的室外道路场景（机器人行走路径约1.2 km，途经行人密集区域、非机动车道和植被密集区域等多个场景，包含行人、非机动车、机动车等动态目标，在场景的不同区域，场景条件可分为稳定环境和复杂环境）进行采集。该数据集包含2048张彩色影像（分辨率为1280×720像素），主要用于测试室外场景复杂环境和稳定环境下视觉定位结果。This dataset is developed for fundamental research on vision-based robotic environmental modeling, localization and navigation. To address the technical bottleneck of real-time high-precision visual localization for robots in complex environments, we collected data using a binocular RGB color camera mounted on a mobile visual robotic platform in outdoor road scenarios on the campus of Tongji University. The mobile robot traversed a path of approximately 1.2 km, passing through diverse scene types including dense pedestrian zones, non-motorized vehicle lanes and densely vegetated areas, with dynamic objects such as pedestrians, non-motorized vehicles and motor vehicles present. The scenario is divided into stable and complex environmental conditions across different regions. The dataset contains 2048 color images with a resolution of 1280×720 pixels, and is primarily used to test visual localization results in both complex and stable outdoor environments.

搜集汇总

背景与挑战

背景概述

该数据集面向机器人环境建模与定位导航研究，通过移动视觉机器人平台的双目RGB相机在同济大学室外道路场景采集，包含约1.2公里路径的2048张彩色影像，用于测试复杂和稳定环境下的视觉定位性能。数据涵盖行人、车辆等动态目标及多种场景条件，旨在支持视觉定位技术突破。