Research on Image Semantic Information Mining Based On Latent Dirichlet Allocation Model

Focusing on the issues of lacking of semantic description on image identification and methods of mapping from low-level semantics to high-level semantics, this paper describes the experiments of identification of image semantic information by using LDA model, which can achieve the mapping from image visual feature to high-level semantics, and experiments on the data sets of Corel 5k and Corel 30k. The experiments test and verify that LDA model performs a good stability on identification of image semantic information, and advantageous on dimension accuracy and recall rate, which provide a new solution and an embodiment of the identifying of image semantics intelligently and automatically.