As one sort of cultural relic, glassy antiques can effectively convey the historical information of a certain era and reveal the cultural exchanges in different regions as a symbol of foreign trade. However, due to long-term weathering and corrosion when buried in soil, the shape, color, and chemical components of a glassy antique can change considerably, and hence identifying it and recognizing its category is particularly difficult. Clustering is a popular technique of data analysis and data mining. K-means is one of the most popular data mining algorithms, as it is simple, scalable, and easy to modify in different contexts and fields of application. This paper uses k-means to find the clustering center showing the characteristics of the chemical components of different categories of glassy antiques. Then the particle swarm optimization algorithm (PSO) that offers a globalized detailed search methodology is utilized to improve the k-means clustering. The new result compared to the previous one of traditional k-means clusterin shows better classification capacity. Finally, it compares the results of k-means with that of PSO-k-means and analysis the advantages and disadvantages of PSO-k-means.