Near-collision
收藏资源简介:
本研究贡献了一个名为Near-collision的大型数据集,包含13,658个以第一人称视角拍摄的室内视频片段,旨在为移动机器人提供更直观的碰撞预测数据。数据集中的每个视频片段至少展示了一个人的轨迹,最终与装有摄像头的移动行李箱形平台接近(即接近碰撞)。数据集的创建过程涉及使用立体摄像机和LIDAR传感器进行数据收集和标注,确保了数据的准确性和可靠性。该数据集主要应用于预测移动机器人与附近行人之间的接近碰撞时间,为动态路径规划中的碰撞避免提供支持。
This study contributes a large-scale dataset named Near-collision, which consists of 13,658 indoor video clips captured from a first-person perspective. The dataset is designed to provide more intuitive collision prediction data for mobile robots. Each video clip in the dataset depicts the trajectory of at least one person who ultimately comes close to a camera-equipped mobile luggage-shaped platform (i.e., near-collision). The dataset creation process involves data collection and annotation using stereo cameras and LIDAR sensors, ensuring the accuracy and reliability of the data. This dataset is primarily applied to predict the near-collision time between mobile robots and nearby pedestrians, providing support for collision avoidance in dynamic path planning.




