Texture Image Recognition Based On Evolution Probabilistic Neural Network

Along with the continuous deepening of computer application, it is hoped that the computer can simulate all kinds of human activities. Accordingly, effectively assist in the production and human life. Computer vision is one of the important tasks to complete this goal .Its purpose is to imitate the human eye to apperceive and cognize the outside world .The texture of universality, as well as the important role played by it in the human perception and cognitive process to the outside world Therefore, the study to texture plays an very important role in computer vision.To improve the accuracy of texture image recognition, the differential evolution method is introduced to make up the shortage of basic probabilistic neural network on the basis of basic probabilistic neural network has the higher recognition when it is applied to classification. Consequently, an evolution probabilistic neural network is proposed in this paper and it is applied into texture image recognition. The differentia evolution algorithm greatly quickens the convergence speed by constructing the parameter vector directly with real number and choosing the initial community randomly with the unified probability distribution; In the algorithm, the diversity of community can be easily maintained because of broad overlapping operation (adopt multi-component of individual randomly); It is sure to get the optimal value from the community because the choose operation has pertinence. The experiment result indicates: The evolution probabilistic neural network has the higher recognition accuracy and faster convergence speed.By analyzing the feature of texture images found out, in the actual texture recognition, some types of texture feature is more obvious, and only require fewer feature parameters can be identified, and some types of texture feature is inconspicuous that need more feature parameters as input parameters can arrive at the higher recognition accuracy. A double evolution probabilistic neural network is put forward and it is applied into texture image recognition in this paper to solve the above problem, which is based on the evolution probabilistic neural network. This paper have done a lot of experiments and proposed an index which can effective measure whether the texture image feature is obvious, according to the local larger energy pixels representing the obvious feature in the original image. Consequently, it ensures the effectiveness of double evolution probabilistic neural network. The application instance indicates that using the double evolution probabilistic neural network to carry out texture image recognition not only has the higher recognition accuracy, but also has the faster recognition speed.