Automated Statistical Methods for Analyzing Color and Pattern Characteristics in Large-Scale Architectural Heritage: A Case Study of Brick-and-Stone Sections of the Ming Great Wall

The colors of cultural heritage are visual characteristics that embody historical and aesthetic values; however, traditional architectural heritage color analysis depends on manual field surveys, limiting efficiency and scalability. Consequently, large-scale heritage sites such as the Ming Great Wall have been perceived as uniformly “gray”, and their artistic value has been underestimated. To address this limitation, this study develops an automated method for analyzing large-scale architectural heritage colors through semantic segmentation and image-based color statistics, using the brick-and-stone sections of the Ming Great Wall in Beijing, Tianjin, and eastern Hebei. A total of 478,143 UAV facade images were collected, and a semantic segmentation dataset was constructed. A training-optimized Mask2Former model was used to segment material components, while K-means clustering and the HSV color model were applied to extract and visualize dominant colors. The model achieved an mIoU of 80.15% and accurately segmented brick and stone materials. The results show that saturated stone materials are predominantly distributed on hollow defense towers. Field verification identified four stone color composition patterns, suggesting links among lithology, construction choices, and component-level color organization. These findings challenge the gray-wall perception of the Ming Great Wall and provide a transferable framework for heritage color analysis, monitoring, and conservation.