An Empirical Study on the Accuracy of Automated Segmentation of Single Characters in Wuwei Hanjian Based on FastSAM

This study investigates the performance of FastSAM (Fast Segment Anything Model) in automated single-character segmentation for Wuwei Han Slips images, particularly under real-world conditions without image preprocessing. Experimental results demonstrate that FastSAM achieves segmentation accuracies of 99.2%, 84.0%, and 78.7% for the King's Staff Ten Bamboo Slips, Wuwei Ritual Bamboo Slips, and Wuwei Medical Bamboo Slips, respectively, with corresponding leakage rates of 0%, 7.2%, and 5%. This research not only establishes baseline data for digital segmentation techniques in cultural heritage preservation but also reveals the strengths and limitations of FastSAM in ancient text scenarios, providing valuable technical insights for future digital conservation efforts.