万集路侧交通参与者轨迹数据集

万集交通轨迹数据集来源于真实交通场景，融合了多路路侧激光雷达融合点云数据和相机视频数据，对真实交通场景中每一个交通参与者运动过程的三维信息进行时序数据采样。针对连续帧点云数据，标注包括目标ID，采样时间，目标位置，形状大小，航向角，交通信号灯状态等信息。 轨迹数据集信息轨迹数据信息主要包括目标基本信息、车辆驾驶行为、行人/非机动车行为、交通参与者交互行为：1、目标基本信息主要包括目标id、类别、航向角、速度、加速度、相对坐标等；2、车辆驾驶行为包括停止、直行、左转、右转、变道、掉头等；3、行人/非机动车行为包括停止、运动等；4、交通参与者交互行为主要有车车交互、人车交互等，包括直行-右转碰撞、邻车直行碰撞、对向直行碰撞、鬼探头碰撞、左转-直行碰撞等。The Wanji Traffic Trajectory Dataset originates from real-world traffic scenarios. It integrates multi-channel roadside LiDAR point cloud data and camera video data, and conducts temporal data sampling on the 3D motion information of each traffic participant in real traffic scenarios. For consecutive frame point cloud data, the annotations cover target ID, sampling timestamp, target position, shape and size, heading angle, traffic light status, and other relevant information. The trajectory dataset primarily includes four categories of information: basic target information, vehicle driving behaviors, pedestrian/non-motor vehicle behaviors, and interactive behaviors among traffic participants: 1. Basic target information mainly includes target ID, category, heading angle, velocity, acceleration, relative coordinates, and other related details; 2. Vehicle driving behaviors include stopping, going straight, turning left, turning right, lane changing, U-turn, and other actions; 3. Pedestrian/non-motor vehicle behaviors include stopping, moving, and other states; 4. Interactive behaviors among traffic participants mainly involve vehicle-to-vehicle interaction and vehicle-to-pedestrian interaction, such as straight-right turn collision, adjacent vehicle straight collision, oncoming straight collision, ghost pedestrian collision, left turn-straight collision, and other typical interaction scenarios.

搜集汇总

背景与挑战

背景概述

该数据集是一个来源于真实交通场景的路侧交通参与者轨迹数据集，融合了多路激光雷达点云和相机视频数据，提供了交通参与者运动的三维时序信息。其标注信息非常详尽，不仅包括目标的基本属性（如ID、类别、位置、速度），还细致分类了车辆驾驶行为（如直行、转弯、变道）、行人/非机动车行为以及交通参与者之间的多种典型交互行为（如各类碰撞场景），适用于智能交通和自动驾驶领域的相关研究。