ViNU: Vision-Navigation Dataset for Unstructured Environment
收藏资源简介：
The dataset was initially gathered using the Robot Pomona, equipped with an RGBD camera, lidar, IMU, and GPS. This diverse sensor setup allowed for the collection of color images, depth images, GNSS data, point clouds, and IMU data. Initially, all data were captured and stored in a ROS bag format. Subsequent to the initial collection, the data were synchronized offline and transferred to a new ROS bag to ensure accurate alignment across the different data types. Finally, the synchronized data were organized and extracted into separate folders for each data type, resulting in distinct datasets for color images, depth images, GNSS, point clouds, and IMU data. This structured organization facilitates easier access and use for research and development in navigating unstructured environments. Potential Applications: Autonomous Navigation: Enhancing the ability of autonomous vehicles to navigate in complex, unstructured environments such as off-road terrains or disaster-stricken areas. Robotics Research: Providing a rich source of sensor data for developing and testing algorithms related to simultaneous localization and mapping (SLAM), object recognition, and path planning. Augmented Reality: Supporting AR applications that require real-time environmental mapping and interaction. Environmental Monitoring: Assisting in tasks that involve monitoring changes in environments over time, useful in ecological research or urban development planning. Machine Learning: Serving as a training and validation dataset for machine learning models focused on sensor fusion, depth perception, and predictive analytics in dynamic scenarios.
本数据集最初由搭载RGBD相机（RGBD Camera）、激光雷达（LiDAR）、惯性测量单元（IMU）与全球定位系统（GPS）的Pomona机器人采集得到。该多样化的传感器配置可采集彩色图像、深度图像、全球导航卫星系统（GNSS）数据、点云数据以及IMU数据。 采集初期，所有数据均以ROS（Robot Operating System）包格式存储。初始采集完成后，研究人员对数据进行离线同步，并将同步后的数据存入新的ROS包中，以确保不同类型数据间的精准对齐。最终，研究人员将同步后的数据按类型整理并提取至各自独立的文件夹中，由此得到彩色图像、深度图像、GNSS、点云以及IMU五类独立数据集。 这种结构化的组织方式，可为非结构化环境导航相关的研发工作提供更便捷的数据获取与使用途径。 潜在应用场景： 自主导航：提升自动驾驶车辆在复杂非结构化环境（如越野地形或灾区）中的导航能力。 机器人学研究：为开发与测试同步定位与建图（SLAM）、目标识别以及路径规划相关的算法提供丰富的传感器数据源。 增强现实（AR）：为需要实时环境建模与交互的AR应用提供支持。 环境监测：辅助开展长期环境变化监测任务，可应用于生态研究或城市发展规划领域。 机器学习：作为训练与验证数据集，用于聚焦动态场景下传感器融合、深度感知以及预测分析的机器学习模型开发。
