Threshold selection for detecting Swiss Cheese polar patterns on Mars using orbital images

Abstract: Threshold selection plays a crucial role in detecting complex and irregular surface features, such as the Swiss Cheese formations found in the south polar region of Mars. This study aims to evaluate and compare the performance of manual and automatic thresholding strategies for detecting Swiss Cheese features. The automatic strategies tested include adaptive thresholding (with average and median-based variants), Otsu’s method, and multilevel thresholding, all integrated into a detection workflow based on digital image processing and mathematical morphology. These approaches were applied to orbital images with spatial resolutions of 0.25 m and 1.5 m. Manual thresholding achieved the highest precision (97.59%) and overall quality (83.80%). Among the automated strategies, multilevel thresholding and Otsu’s method yielded the best results, with multilevel thresholding reaching 87.29% precision and 28.38% quality, while Otsu’s method reached 78.39% precision and 30.91% quality. These findings highlight the challenge of defining a global threshold due to illumination variability, contrast differences, and the irregular morphology of the Swiss Cheese formations. The results support the development of a more systematic and reproducible workflow for planetary surface analysis.