光谱成像技术研究室知识库

The Sanxingdui Ruins site, situated in Sichuan Province, China, constitutes a significant component of the ancient Shu culture. Confronted with the challenge of excavating and analyzing cultural relics, this study introduces hyperspectral methods for the first time, utilizing the cultural relics unearthed from sacrificial pits. It proposes a novel algorithm for the extraction and analysis of Sanxingdui cultural relics, while also establishing a comprehensive data acquisition and analysis system. The specific innovations of this research are as follows: 1) Acquiring spectral curves of representative materials such as ivory, jade, bronze, and stone pillars, and constructing a spectral database. 2) Introducing the Dual-encoder UNet (D-Unet) model from a spatial perspective to achieve preliminary extraction of cultural relics. 3) Proposing an unsupervised Fuzzy C-means (FCM) algorithm from a spectral perspective to analyze the material properties of cultural relics. Experimental results demonstrate that the proposed algorithm successfully achieves the extraction and analysis of Sanxingdui cultural relics. Ablation experiments further reveal that the algorithm yields excellent results in scenarios involving exposed cultural relics, semi-exposed cultural relics on the ground, and typical cultural relics. Moreover, the algorithm can be extended to other aspects of archeological excavation and analysis. During the experiment, it was observed that: 1) The cultural relics' surfaces contain distinctive substances, tentatively identified as proteins, suggesting the presence of textile capabilities in the ancient Shu civilization. 2) An unearthed bronze mask exhibits structural asymmetry, contrasting with other highly refined artifacts. Through intelligent synthesis technology analysis, a totem-like target was discovered at the center of the eyebrow, potentially associated with the veneration of eyes in Sanxingdui cultural relics, thereby providing additional validation of the algorithms effectiveness. © 2023 Elsevier Ltd