Skip to content

Camera-Pose Robust Crater Detection from Chang'e 5

Matthew Rodda (Uni Adelaide)Earth 2

As space missions aim to explore increasingly hazardous terrain, crater-based pose estimation (CBPE) may provide the accurate and timely position estimates required to ensure safe navigation. Crater-detection algorithms (CDAs) are a crucial initial step to CBPE, using images collected from a spacecraft's onboard cameras to detect impact craters in the environment. Existing literature has proposed many algorithms for crater detection, however performance has only been evaluated within a narrow operating scenario. In this work, we demonstrate the challenge in detecting craters from images containing off-nadir view angles, using Mask R-CNN, a state of the art detection algorithm. We demonstrate training on real-lunar images is superior to a simulated dataset despite lacking training images containing off-nadir view angles, achieving detection performance of 63.1% F1-score and ellipse-regression performance of 0.701 intersection over union. This work provides the first quantitative analysis of CDA performance on images containing off-nadir view angles. Towards the development of robust CDAs, we additionally provide the first annotated dataset with off-nadir view angles from the Chang'e 5 Landing Camera, available here: https://zenodo.org/doi/10.5281/zenodo.11326449.

Tue 14:10 - 14:30
Navigation & Control