highD Dataseth (The Highway Drone Dataset Naturalistic Trajectories of 110 500 Vehicles Recorded at German Highways)

highD数据集是在德国高速公路上记录的自然主义车辆轨迹的新数据集。使用无人机，空中透视可以克服已建立的交通数据收集方法 (例如遮挡) 的典型限制。在六个不同的地点记录了交通，其中包括110多辆500车辆。自动提取每辆车的轨迹，包括车辆类型、尺寸和机动。使用最先进的计算机视觉算法，定位误差通常小于10厘米。尽管该数据集是为高度自动化车辆的安全验证而创建的，但它也适用于许多其他任务，例如交通模式分析或驾驶员模型的参数化。

The highD dataset is a novel dataset of naturalistic vehicle trajectories recorded on German highways. Utilizing unmanned aerial vehicles (UAVs), the bird's-eye perspective overcomes typical limitations of established traffic data collection methods such as occlusion. Traffic was recorded at six distinct locations, encompassing more than 110,000 vehicles. Trajectories for each tracked vehicle are automatically extracted, including their type, dimensions, and driving maneuvers. Using state-of-the-art computer vision algorithms, the localization error is typically less than 10 centimeters. Although this dataset was created for the safety validation of highly automated vehicles, it is also applicable to numerous other tasks such as traffic pattern analysis or parameterization of driver models.

搜集汇总

背景与挑战

背景概述

highD数据集是一个在德国高速公路上使用无人机采集的自然主义车辆轨迹数据集，记录了超过110,500辆车的轨迹信息，包括车辆类型、尺寸和机动数据，定位精度高（误差小于10厘米）。该数据集最初为高度自动化车辆的安全验证而设计，但也广泛适用于交通模式分析、驾驶员模型参数化等多种研究任务。