Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LuSNAR:A Lunar Segmentation, Navigation and Reconstruction Dataset based on Muti-sensor for Autonomous Exploration

Jiayi Liu, Qianyu Zhang, Xue Wan, Shengyang Zhang, Yaolin Tian, Haodong Han, Yutao Zhao, Baichuan Liu, Zeyuan Zhao, Xubo Luo

TL;DR

LuSNAR addresses the lack of diverse, high-precision lunar data for autonomous perception and navigation by introducing a multi-task, multi-scene, multi-sensor benchmark generated in Unreal Engine. It supports semantic segmentation, SLAM, and 3D reconstruction with rich ground-truth labels (depth, 2D/3D semantics, pose) across nine varied lunar scenes and multiple sensor modalities. Through comprehensive experiments, LuSNAR demonstrates the feasibility of joint evaluation and highlights the strengths and limitations of current state-of-the-art methods in lunar contexts. The dataset is poised to enable standardized ground verification and accelerate autonomous lunar rover development, with future work to broaden scene coverage and extend tasks such as multi-sensor fusion and orbital-alignment positioning.

Abstract

With the complexity of lunar exploration missions, the moon needs to have a higher level of autonomy. Environmental perception and navigation algorithms are the foundation for lunar rovers to achieve autonomous exploration. The development and verification of algorithms require highly reliable data support. Most of the existing lunar datasets are targeted at a single task, lacking diverse scenes and high-precision ground truth labels. To address this issue, we propose a multi-task, multi-scene, and multi-label lunar benchmark dataset LuSNAR. This dataset can be used for comprehensive evaluation of autonomous perception and navigation systems, including high-resolution stereo image pairs, panoramic semantic labels, dense depth maps, LiDAR point clouds, and the position of rover. In order to provide richer scene data, we built 9 lunar simulation scenes based on Unreal Engine. Each scene is divided according to topographic relief and the density of objects. To verify the usability of the dataset, we evaluated and analyzed the algorithms of semantic segmentation, 3D reconstruction, and autonomous navigation. The experiment results prove that the dataset proposed in this paper can be used for ground verification of tasks such as autonomous environment perception and navigation, and provides a lunar benchmark dataset for testing the accessibility of algorithm metrics. We make LuSNAR publicly available at: https://github.com/zqyu9/LuSNAR-dataset.

LuSNAR:A Lunar Segmentation, Navigation and Reconstruction Dataset based on Muti-sensor for Autonomous Exploration

TL;DR

LuSNAR addresses the lack of diverse, high-precision lunar data for autonomous perception and navigation by introducing a multi-task, multi-scene, multi-sensor benchmark generated in Unreal Engine. It supports semantic segmentation, SLAM, and 3D reconstruction with rich ground-truth labels (depth, 2D/3D semantics, pose) across nine varied lunar scenes and multiple sensor modalities. Through comprehensive experiments, LuSNAR demonstrates the feasibility of joint evaluation and highlights the strengths and limitations of current state-of-the-art methods in lunar contexts. The dataset is poised to enable standardized ground verification and accelerate autonomous lunar rover development, with future work to broaden scene coverage and extend tasks such as multi-sensor fusion and orbital-alignment positioning.

Abstract

With the complexity of lunar exploration missions, the moon needs to have a higher level of autonomy. Environmental perception and navigation algorithms are the foundation for lunar rovers to achieve autonomous exploration. The development and verification of algorithms require highly reliable data support. Most of the existing lunar datasets are targeted at a single task, lacking diverse scenes and high-precision ground truth labels. To address this issue, we propose a multi-task, multi-scene, and multi-label lunar benchmark dataset LuSNAR. This dataset can be used for comprehensive evaluation of autonomous perception and navigation systems, including high-resolution stereo image pairs, panoramic semantic labels, dense depth maps, LiDAR point clouds, and the position of rover. In order to provide richer scene data, we built 9 lunar simulation scenes based on Unreal Engine. Each scene is divided according to topographic relief and the density of objects. To verify the usability of the dataset, we evaluated and analyzed the algorithms of semantic segmentation, 3D reconstruction, and autonomous navigation. The experiment results prove that the dataset proposed in this paper can be used for ground verification of tasks such as autonomous environment perception and navigation, and provides a lunar benchmark dataset for testing the accessibility of algorithm metrics. We make LuSNAR publicly available at: https://github.com/zqyu9/LuSNAR-dataset.
Paper Structure (16 sections, 10 equations, 13 figures, 11 tables)

This paper contains 16 sections, 10 equations, 13 figures, 11 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Planetary Semantic segmentation datasets
  4. Navigation datasets
  5. Dataset features
  6. Multi-task supported
  7. Semantic perception
  8. SLAM
  9. 3D Reconstruction
  10. Diversity in lunar topographic relief and objects density
  11. multi-label
  12. Experiment
  13. Semantic Segmentation
  14. SLAM
  15. 3D Reconstruction
  16. ...and 1 more sections

Figures (13)

  • Figure 1: Overview of 9 lunar surface scenes with different topographic relief and density of objects.
  • Figure 2: Qualitative and quantitative comparison of the 9 scenes. (a) The topographic relief and density of objects in 9 scenes. (b) The topographic relief trends of the 9 scenes. (c) The abundance of rocks in 9 scenes. (d) The statistics on the number of mountain ranges and impact craters in 9 scenes.
  • Figure 3: Simulation data generation pipeline.
  • Figure 4: The statistical distribution of semantic labels in the LuSNAR dataset. (a) The proportion of different object quantities in images across all scenes. (b) The proportion of different object quantities in LiDAR point clouds across all scenes. (c) The proportion of object quantities in images across 9 different scenes. (d) The proportion of object quantities in LiDAR point clouds across 9 different scenes.
  • Figure 5: Extrinsic settings for rover and sensors.
  • ...and 8 more figures