Bridging the Data Gap of Asteroid Exploration: OAISYS Extension for Synthetic Asteroids Creation

Kurzfassung

In this work, we address the limited availability of real asteroid data for visual applications in the context of asteroid exploration and research. We propose an extension for the photo-realistic simulation environment OAISYS (Outdoor Artificial Intelligence SYstems Simulator) with a dedicated asteroid plugin, designed to address the unique challenges of asteroid missions. This plugin enables the generation of realistic asteroid models by deforming generic shape models and placing them on randomized, plausible trajectories. Additionally, it incorporates advanced visual crater modeling to simulate the often irregular surfaces of asteroids. The plugin generates photo-realistic training data and offers hierarchical pixel-wise annotations down to the instance level for each crater, making it highly useful for tasks such as crater detection, mapping, and counting. By training a detector on a dataset rendered with our OAISYS extension we investigate the plugin's sim-to-real gap and applicability on real-world asteroid imagery for applications such as crater detection and counting. The results indicate that our OAISYS extension provides a valuable resource for the development and testing of future automated asteroid exploration technologies, particularly in autonomous navigation and surface analysis systems. Code will be released at https://github.com/DLR-RM/oaisys.