Facial Complexion Classification of Traditional Chinese Medicine Based on Statistical Features and MobileViT

Traditional Chinese medicine believes that complexion is an external reflection of the changes of human diseases, and the automatic classification of complexion is one of the important directions of the objectification of TCM face diagnosis. At present, the traditional machine learning method is still the mainstream method in the field of face diagnosis of traditional Chinese medicine, and the application of deep learning method is still relatively few. Based on the perspective of feature fusion, this paper proposes a TCM facial complexion classification method based on statistical features and MobileViT. Firstly, the detection and segmentation of facial skin is done on the sample image, and then multiple color features, texture feature and luster feature of the face are extracted. In addition, the MobileViT model is used as a feature extractor to extract the self-learning features of the face. Finally, the statistical features are fused with the self-learning features, and the classification test is performed using three classifiers: LR, LDA, and MLP. The results show that the accuracy of face complexion classification of the method proposed in this paper reaches 94.1243%, which is significantly improved compared with the use of statistical features and MobileViT model alone.