TOMD
收藏资源简介:
TOMD是一个为狭窄和非结构化的类似路径环境设计的综合数据集。该数据集包含高保真的多模态传感器数据,包括128通道激光雷达、立体图像、GNSS、IMU和光照测量。数据是在不同的环境条件下通过重复运行收集的。此外,我们还提出了一种新的动态多尺度数据融合模型,用于在类似路径区域中精确预测可通行路径。该研究调查了在不同光照水平下,各种融合过程(早期、交叉和混合)对模型性能的影响,包括低光、正常环境光照和明亮条件。结果表明,我们的方法有效,性能随光照水平变化而变化,并且数据集在不同环境条件下具有潜在的应用性。我们的工作为推进基于路径的越野导航提供了一个宝贵的资源,并且我们公开发布了TOMD,以建立未来该研究领域的基准。
TOMD is a comprehensive dataset designed for narrow and unstructured path-like environments. It includes high-fidelity multi-modal sensor data, namely 128-channel LiDAR, stereo images, GNSS, IMU, and light intensity measurements. The data was collected via repeated runs under diverse environmental conditions. Furthermore, we propose a novel dynamic multi-scale data fusion model for accurately predicting traversable paths in path-like regions. This study examines the impact of various fusion processes (early, cross, and hybrid) on model performance across different light levels, including low-light, normal ambient lighting, and bright conditions. The results show that our method is effective, with performance varying alongside light levels, and the dataset has potential applicability across varied environmental conditions. Our work offers a valuable resource for advancing path-based off-road navigation, and we have publicly released TOMD to establish a benchmark for future research in this field.

- 1TOMD: A Trail-based Off-road Multimodal Dataset for Traversable Pathway Segmentation under Challenging Illumination Conditions英国达勒姆大学计算机科学系、英国国王学院工程系 · 2025年



