BorealHDR

BorealHDR数据集是由北方机器人实验室创建，专注于高动态范围（HDR）场景下的视觉里程计（VO）性能评估。该数据集包含55条轨迹，总长超过10公里，覆盖多种季节和光照条件，特别在魁北克森林和拉瓦尔大学校园进行采集。数据集不仅包括多曝光立体图像，还整合了3D激光雷达扫描、惯性测量和全球导航卫星系统（GNSS）数据，为每帧图像提供姿态估计。BorealHDR数据集的创建旨在解决现有自动曝光（AE）算法比较的不可重复性问题，通过提供真实世界的图像数据，支持离线基准测试，加速未来算法的发展。

The BorealHDR Dataset was developed by the Northern Robotics Laboratory, dedicated to evaluating the performance of visual odometry (VO) in high dynamic range (HDR) scenarios. It comprises 55 trajectories with a total length exceeding 10 kilometers, covering various seasons and lighting conditions, and was specifically collected in the Quebec forests and the campus of Laval University. The dataset not only includes multi-exposure stereo images, but also integrates 3D LiDAR scans, inertial measurement data, and Global Navigation Satellite System (GNSS) data, which provide pose estimation for each image frame. The development of the BorealHDR Dataset aims to resolve the non-reproducibility issue encountered in existing comparisons of automatic exposure (AE) algorithms. By providing real-world image data, it supports offline benchmarking and accelerates the advancement of future algorithms.