The POLAR Traverse Dataset: A Dataset of Stereo Camera Images Simulating Traverses across Lunar Polar Terrain under Extreme Lighting Conditions

TL;DR

The paper introduces the POLAR Traverse Dataset, a high-fidelity collection of stereo image pairs captured to simulate straight-line traverses over lunar-like polar terrain under extreme lighting. It details a lab-based setup with a lunar regolith simulant, four craters, and low-angle illumination, paired with ground-truth LiDAR scans and pose estimates refined by COLMAP. The dataset comprises 3,960 stereo pairs across 24 traverses, with multiple camera heights, pitches, and exposure times to probe perception under challenging lighting. It analyzes COLMAP's performance under these conditions, highlighting both feasibility with careful tuning and limitations due to the lab environment, and it provides access for developing robust visual odometry and stereo-vision algorithms for VIPER and Artemis-style missions. Overall, the work offers a valuable benchmark for perception research in lunar polar scenarios and documents practical considerations and limitations of lab-based simulations.

Abstract

We present the POLAR Traverse Dataset: a dataset of high-fidelity stereo pair images of lunar-like terrain under polar lighting conditions designed to simulate a straight-line traverse. Images from individual traverses with different camera heights and pitches were recorded at 1 m intervals by moving a suspended stereo bar across a test bed filled with regolith simulant and shaped to mimic lunar south polar terrain. Ground truth geometry and camera position information was also recorded. This dataset is intended for developing and testing software algorithms that rely on stereo or monocular camera images, such as visual odometry, for use in the lunar polar environment, as well as to provide insight into the expected lighting conditions in lunar polar regions.

Figures (5)

• Figure 1: Hardware setup extended over SSERVI test bed with lunar terrain and lighting.
• Figure 2: Closeup of stereo bar attached to end of jib arm with pitch mount (blue).
• Figure 3: Top-down view of the scene inside the test bed from the processed ground truth scan, displaying lunar-like terrain and the human-made object (lower left corner). Color denotes height (blue is lowest, red is highest). Approximate light locations and camera positions are shown for each viewing direction (black is forward, grey is reverse).
• Figure 4: Example left camera images from the first sampling location for the two forward-facing terrain views (1, 2) available in the POLAR Traverse Dataset. Both images were taken with the same extrinsic parameters (14° camera pitch, 1.3 m above the terrain, 50 ms exposure time).
• Figure 5: Dense reconstruction from COLMAP generated using images from terrain view 1. The scene reconstruction contains a number of points from the scene outside the test bed.