📄 中文摘要
低光照条件对自主驾驶中的视觉感知系统构成了挑战。提出了一个新的基准数据集（命名为DarkDriving），旨在研究低光增强技术。现有的真实世界低光增强基准数据集通常仅在小范围和静态场景中通过控制不同曝光来收集。当前夜间驾驶数据集中的暗图像与精确对齐的白天图像缺乏对应关系。在动态驾驶场景中收集真实世界的昼夜对齐数据集的极大困难，显著限制了该领域的研究。通过提出的自动昼夜轨迹跟踪方法，解决了这一问题，推动了低光照条件下自主驾驶技术的发展。
暗驾驶：用于黑暗环境中自主驾驶的真实世界昼夜对齐数据集
📄 English Summary
DarkDriving: A Real-World Day and Night Aligned Dataset for Autonomous Driving in the Dark Environment
Low-light conditions pose significant challenges to vision-centric perception systems for autonomous driving in dark environments. A new benchmark dataset, named DarkDriving, is proposed to investigate low-light enhancement techniques for autonomous driving. Existing real-world low-light enhancement benchmark datasets are typically collected by controlling various exposures only in small ranges and static scenes. The dark images in current nighttime driving datasets lack precisely aligned daytime counterparts. The extreme difficulty in collecting a real-world day and night aligned dataset in dynamic driving scenes has significantly limited research in this area. An automatic day-night trajectory tracking method is proposed to address this issue, advancing the development of autonomous driving technologies under low-light conditions.