Scene classification based on block latent semantic
基于分块潜在语义的场景分类方法

A novel scene classification method was presented based on block latent semantic. The image blocks were first extracted on a regular grid and the visual words in blocks were used to describe every block, and then block latent semantic models were achieved by using Probabilistic Latent Semantic Analysis (PLSA). The latent semantic model was used to find the latent semantic in image block and their spatial distribute in image. Finally, this feature was used to construct a SVM model to classify scene. Experimental results show that this method has satisfactory classification performances on a large set of 13 categories of complex scenes.