High-Accuracy Crater Detection for Complex Planetary Surfaces

Impact craters play a crucial role in planetary science, as they serve as windows into the geological evolution of planetary surfaces and are key considerations in selecting landing sites. However, most existing crater detection algorithms are either closed source or difficult to use due to complex configurations, and many are limited to specific celestial bodies, performing well only on simple, flat surfaces. To address these limitations, we present the “automatic crater detection (ACD) tool,” a user-friendly software solution built on deep learning algorithm that was selected through experimental parameter comparisons. By incorporating active learning techniques, the tool continually improves through the inclusion of samples from varying landforms. This allows for accurate crater detection on complex and diverse planetary surfaces, with a minimum detectable crater diameter of 10 pixels. Tested on standard image datasets from the Moon, Mars, and Mercury, the method achieved high accuracy, achieving ${F}1$ -scores of 92.8%, 91.7%, and 92.2% for the Moon, Mars, and Mercury, respectively. The ACD tool is wrapped with an interactive user interface that ensures easy utilization, allowing users to generate ready-to-use output data with a simple click. The introduction of ACD tool can empower scientists with a straightforward yet powerful tool for conducting crater research more efficiently, supporting tasks such as surface age estimation and landing site safety analysis. The software can be downloaded via the following link: https://www.scidb.cn/s/yemuYb.