Abstract: The article seek to discover the object categories' semantic probabilistic model based on statistical text analysis and we applied this new model on object recognition and Scene analysis. First, the image was represented by a set composed of local feature regions. Then, found the probability among image, local regions and semantic category based on the new model helps to calculate the posterios and recognizing the object. EM algorithm was used to estimate the parameters of the model. Experiments show the good performance on object recognition in the cluster background and also show the feasibility of the scene analysis.

Key words: Category recognition, Scene analysis, Semantic probabilistic model, Graph model