Research on Similarity Matching of Dazu Rock Carvings in China Based on a Trimodal Fusion of pHash, SIFT, and MSRoVGG

A novel quantitative model for assessing the similarity of cultural relics, such as China's Dazu Rock Carvings from the Song Dynasty, achieves 100% recognition accuracy for identical statue types within the same niche while effectively distinguishing between different iconographic types. This breakthrough addresses the limitations of traditional qualitative expert assessments by integrating a trimodal feature fusion approach: Perceptual Hash (pHash) for global structural screening, SIFT for high-dimensional texture feature extraction and local keypoint matching, and an improved MSRoVGG deep learning architecture for semantic-level verification. The model demonstrates robust capabilities in typological classification and stylistic differentiation, capturing nuanced variations in artistic styles and vestimentary features between Buddhist and secular figures.

Xue Yang, & Jie Tang. (2026). Research on Similarity Matching of Dazu Rock Carvings in China Based on a Trimodal Fusion of pHash, SIFT, and MSRoVGG. Zenodo. https://doi.org/10.5281/ZENODO.19847401