Object perception model in visual cortex based on Bayesian network

Motivating from biological visual cues in the cortex, by simulating visual information processing and transmission mechanism in the human brain, and using Bayesian network to design object perception model in the visual cortex, this paper proposed an object perception model based on Bayesian network. First, extracted shape feature, color feature, texture feature of the given images; Second, normalized these features and inputed them all to Bayesian network for inference and learning; Third, carried out two experiments to test the validity and reliability of the proposed model. Experiment results shown that the proposed model is reasonable and robust, can integrate all possible information and combine varieties of evidence to implement uncertainty inference, can solve problems with uncertainty and incomplete effectively. The proposed model achieved better recognition performance on the given experimental image datasets, obtained a higher recognition accuracy compared with other methods, and better solved various of recognition difficulties in visual object recognition.