KITTI 数据集的住宅序列

本数据集源自KITTI数据集的住宅序列，由米兰比可卡大学信息、系统和通信系创建，旨在为交叉路口的视觉定位提供基准。数据集包含48个车辆接近交叉路口的子序列，每个子序列包含双目视频和车辆位置的精确估计。数据集创建过程中，研究团队评估了多个公开数据集，最终选择KITTI数据集的住宅序列，因其包含双目摄像机和GPS-RTK数据。尽管数据集仅涵盖德国某城市的交叉路口，但其包含了不同几何形状、光照和交通条件的交叉口。该数据集主要用于评估和比较视觉定位技术在交叉路口的应用效果，特别是在车辆接近交叉口时的定位精度。

This dataset is derived from the residential sequences of the KITTI dataset, and was developed by the Department of Information, Systems and Communications of the University of Milan-Bicocca, with the aim of providing a benchmark for visual localization at intersections. The dataset comprises 48 subsequences in which vehicles approach intersections, with each subsequence containing stereo videos and precise estimates of vehicle positions. During the dataset construction process, the research team evaluated multiple public datasets, and ultimately selected the residential sequences of the KITTI dataset due to their inclusion of stereo cameras and GPS-RTK data. Although the dataset only covers intersections in a city in Germany, it includes intersections with diverse geometries, lighting conditions and traffic scenarios. This dataset is primarily used to evaluate and compare the performance of visual localization technologies applied at intersections, especially the localization accuracy when vehicles are approaching such intersections.