Research on recognition of wood texture based on integrated neural network classifier

This paper regarded indoor environment and decorative lumbers as the research object, and presented an integrated rule on the measure level of the overall recognition rate of the sample based on BP neural network classifier by trying to recognize the position of maximum value of vectors W to identify the categories of the samples. The number of networks could be delimited with the highest prior recognition rate of a single network as the center and the difference between the recognition rates as the radius, which integrated the parameters of 7 BP neural networks. The paper extracted the Gray Level Co-occurrence Matrix parameters of plate texture, and adopted simulated annealing algorithm to select character parameters as an input feature vectors of integrated neural network classifier. Experiment results show that recognition rate of the integrated neural network classifier is superior to single network or the nearest neighborhood classifier, the average recognition ratio of 10 texture samples reaches 90.25%.