An Automatic Measurement Method for Architectural Heritage Based on Point Cloud Semantic Segmentation Algorithm: A Case Study of the Hollow Watchtowers of the Ming Great Wall

Abstract. Dimensional data are critical for the assessment, conservation, and restoration of architectural heritage. Traditional manual measurement methods are time-consuming and labor-intensive, particularly for large, complex, and difficult-to-access structures such as the hollow watchtowers of the Ming Great Wall. This study proposes a new method that combines low-altitude UAV photogrammetry with point cloud semantic segmentation algorithms, using watchtowers along the Ming Great Wall as case studies. First, images were collected using low-altitude UAV, and point cloud data are generated via photogrammetry to capture information of the difficult-to-reach watchtowers located on mountain ridges. Second, a semantic segmentation algorithm was applied to classify different components of the watchtower, and dimensional data were automatically calculated by fitting geometric models. Finally, experimental results showed that this automated method significantly outperforms traditional techniques in both efficiency and accuracy. The overall accuracy of the point cloud semantic segmentation algorithm reaches 90.80, and the error in automatically calculating the length and width of the watchtowers is less than 10%. Through the dimensional analysis of 601 watchtowers, the study identifies dimensional differences among watchtowers under three military jurisdictions during the Ming Dynasty. In summary, this study develops an automatic dimensional analysis method for watchtowers of the Ming Great Wall, improving the efficiency of heritage surveys. The method also shows potential for extension to other large-scale architectural heritage, offering a valuable tool for rapid analysis and digital archiving.