Adverse Weather Kitti 360

New Dataset: Adverse Weather-Augmented Kitti 360

This dataset enhances the KITTI 360 dataset by adding simulated snow, rain, and fog effects to the original clear-weather scans captured by the Velodyne HDL-64 LiDAR sensor. This allows researchers to evaluate their algorithms for tasks like lidar odometry, SLAM, navigation, and 3D object detection under challenging weather conditions.

Scenarios: 9 long-term driving sequences with post-processed GNSS/IMU ground truth localization data.
Data per Scan: Points include x, y, z coordinates, intensity, ring/channel information, and a normalized azimuth angle representing time.
Weather Masks: Each scan includes a mask that identifies points affected by simulated weather (signal attenuation).
Augmented Scans: Three compressed versions (rain, fog, and snow) are provided for each scenario, encoded as binary float32.
Original Scans: Downloadable from the KITTI-360 dataset.
Data Access: A C++/Python code example demonstrates how to read the data format.

This dataset provides a valuable resource for researchers developing robust algorithms that function effectively in adverse weather conditions.

新型数据集：恶劣天气增强型KITTI 360

本数据集通过为Velodyne HDL-64激光雷达（LiDAR）采集的原始晴朗天气扫描点云添加模拟降雪、降雨与雾霭效果，对KITTI 360数据集进行了增强。这使得研究人员能够在复杂恶劣的天气条件下，针对激光雷达里程计、同步定位与地图构建（Simultaneous Localization and Mapping，简称SLAM）、导航以及三维目标检测等任务对其算法进行性能评估。

场景：共包含9段长期驾驶序列，配套经过后处理的全球导航卫星系统（Global Navigation Satellite System，简称GNSS）/惯性测量单元（Inertial Measurement Unit，简称IMU）真值定位数据。

单帧扫描数据：点云包含x、y、z三维坐标、强度值、环/通道信息以及用于表征采集时刻的归一化方位角。

天气掩码：每帧扫描数据均附带掩码，用于标记受模拟天气影响（即信号衰减）的点云。

增强型扫描数据：为每个场景提供三种压缩后的增强版本（降雨、雾霭与降雪），编码格式为二进制float32。

原始扫描数据：可从KITTI 360数据集下载获取。

数据访问：附带C++/Python代码示例，演示如何读取该数据集的格式。

本数据集为研发可在恶劣天气条件下稳定运行的鲁棒算法的研究人员提供了极具价值的研究资源。

搜集汇总

背景与挑战

背景概述

该数据集是基于KITTI 360数据集增强的新资源，通过模拟雪、雨和雾效果添加到原始晴朗天气的LiDAR扫描中，旨在支持恶劣天气条件下的算法评估，如LiDAR里程计、SLAM和3D物体检测。它包含9个驾驶序列，每个扫描点提供坐标、强度等详细信息，并附带天气掩码和压缩版本，为研究人员开发鲁棒算法提供了宝贵工具。