Visual place recognition benchmark for moon and mars like environments
VPR-Space: A Benchmark for Visual Place Recognition in Space Analog Domains
RoboticsComputer Vision and Pattern Recognition
Summary
Recognizing places visually helps robots know where they are by comparing current camera views to past ones, which is especially challenging on the Moon or Mars due to few features. The paper presents a new set of test datasets based on Earth environments similar to these space surfaces. It introduces a better way to decide when two images show the same place, focusing on usefulness for robot navigation instead of just camera positions. The authors tested nine existing methods and found that better solutions are needed for these space-like conditions.
What this means in practice
- •For space robotics teams: Evaluate and improve visual localization systems that help space rovers navigate extraterrestrial-like terrain using the new VPR-Space benchmark datasets.
- •For autonomous vehicle engineers: Test visual place recognition algorithms in feature-poor, challenging outdoor environments that resemble other planets but occur on Earth.
Authors
Richard Fechner, Peer Neubert
Abstract
Visual Place Recognition (VPR) is an important means for loop closure detection in visual Simultaneous Localization and Mapping (SLAM). While state-of-the-art VPR and SLAM approaches excel in urban and feature-rich natural environments, the vast and feature-poor environments of extraterrestrial surfaces like on the Moon or Mars pose severe challenges for existing approaches. This paper presents a novel benchmark, VPR-Space, to evaluate VPR capabilities on extraterrestrial surfaces. There are three contributions: (1) We collected sixteen VPR database-query datasets based on existing data from Space Analog Domains (SAD), terrestrial environments that resemble properties of the Moon or Mars. The collection includes a range of challenges regarding environment, viewpoint change, and illumination conditions. (2) To address the ill-posed nature of VPR, we propose a novel approach to generate the ground-truth information in VPR benchmarks. Instead of defining that a query and datbase image show the same place if and only if their camera poses are within fixed position and viewing direction thresholds, we propose to make the decision dependent on whether the match would be valuable in the downstream SLAM task. To avoid the overhead and restrictions of evaluating VPR in-the-loop within SLAM systems, we provide VPR ground-truth (GT) matrices based on the performance of an ensemble of state-of-the-art Relative Pose Estimators (RPE). Under this GT we require a VPR method to identify exactly those image pairs for which a subsequently applied RPE is expected to provide a valuable pose estimate in a SLAM context. (3) We evaluated nine state-of-the-art VPR approaches. The results show considerable room for improvement and indicate directions for future research. We provide a pre-release version of the dataset on github: $\texttt{https://github.com/rfechner/VPR-Space}$