Sub-Classification and Type Identification of Weathering Silicate Glasses Based on Spectral Clustering

Abstract

Ancient glass products are often weathered to varying degrees after excavation. Detection and analysis of their chemical composition will help to study the ancient glass making process. In this paper, we take a group of ancient glass products as samples, and study the sub-classification and type identification of weathered silicate glasses. Firstly, we identify the main factor of glass weathering through chi-squared test and Fisher’s exact test, then conduct a significance test on the basis of preprocessing various data of glass cultural relics. Secondly, the glass samples are classified into high potassium glass and lead barium glass by spectral clustering method, and further sub-classified into eight classes. The sub-classes are named based on feature extraction, and the clustering quality is tested by contour coefficient. Finally, we analyze the chemical composition affecting the glass types by combining the non-parametric tests to identify the types of glass cultural relics by multiple linear regression with significance testing. In addition, correlation test, spectral clustering and multiple linear regression are used to study the sub-classification and type identification of weathered silicate glasses, which provides a reference for further study of ancient glass making process and weathering process.