Research on ancient ceramic identification by artificial intelligence

Abstract Ancient ceramics occupy a pivotal position in the inheritance of traditional Chinese culture, which is an important carrier of history. The accurate identification of ancient ceramics is a difficult problem that is associated with a great deal of confusion. There are currently two methods used to identify ancient ceramics. The first is the expert visual method, where experienced experts provide recognition and conclusion through the personally empirical combination of vision and tactile sense. The second is the science and technology identification method, which involves analyzing and comparing the composition of ancient ceramics with standard databases, and is more reliable and reasonable for a business environment. Here, artificial intelligence-aided identification of ancient ceramics was investigated. At present, artificial intelligence is widely used in many fields, such as piezoelectric material discovery and material design. Machines are used to simulate human thinking so that problems can be solved in complex situations. In this study, three main ancient visual features of ceramics were transformed into machine vision features: shape as an object contour, ornamentation as an image texture, and inscription as handwritten Chinese characters. Machines instead of experts were able to visually identify ancient ceramics. With this method, when parts of ancient ceramics were damaged, the shape similarity was computed to be higher than 95%, the Euclidean distance between ornamentation was 0.5848, and the inscription recognition rate was 83.33%. These results indicate that this machine-aided method utilizing artificial intelligence is effective for identification of ancient Chinese ceramics.