Multi-label Constitution Identification Based on Tongue Image in Traditional Chinese Medicine

Constitution identification is an important research topic in Traditional Chinese Medicine (TCM), which is the basis of health management and disease prevention. Tongue diagnosis is one of the common methods in TCM treatment. Combining knowledge of Traditional Chinese Medicine and computer technology to analyze tongue images can identify constitution objectively and effectively. The existing intelligent constitution identification algorithms can give only one constitution result for each tongue image. However, in the actual process of constitution identification, individuals often present two or more constitutions. In this paper, a multi-label classification method is used to identify the composite constitution of tongue images, and the feature extraction performance of four different networks is compared. The idea of complex perception is applied to the DenseNet-121 network, and the applicable condition of the classification method based on complex perception is given through experiments.