Terracotta warriors and horses in the mausoleum of the first emperor of the Qin Dynasty has been a focus of scholars’ research since its discovery. Many experts have done qualitative research on the facial features of terracotta warriors and horses, but there is a lack of quantitative evidence. In this study, the face recognition technology was used to quantitatively analyze the facial features of terracotta warriors. The analysis process included image preprocessing, face detection and screening, face feature extraction, face feature vector representation, face feature vector similarity calculation and face feature vector similarity analysis, etc. The facial features of the terracotta warriors were analyzed intuitively through the facial similarity Top-N of a single terracotta warrior. By analyzing the facial similarities of several terracotta warriors through thermal maps, it was found that ten terracotta warriors had “distinctive faces” and ten terracotta warriors had “common faces.” A histogram was used to visualize the facial similarity distribution of terracotta warriors, and a normal distribution law of facial similarity distribution of terracotta warriors was obtained. For the research on terracotta warriors, the face recognition technology provides a quantitative analysis idea, which can not only be used for the analysis of facial features and clothing texture of terracotta warriors, but also lays down a technological foundation for the quantitative analysis of facial features of larger-scale terracotta warriors.