Research on Classification of Wood Surface Texture based on Feature Level Data Fusion

摘要

In order to enhance the precision of wood texture recognition, a kind of wood surface texture recognition method based on feature level data fusion is proposed, which uses GLCM, GMRF and wavelet multi-resolution fractal dimension. First, feature parameters of 3 sorts of wood textures were selected by Simulated Annealing Algorithm, and extracted features which were fatal to image recognition to classify. Next, 3 sorts of texture features were fused on the feature level. With the fused features, the recognition rate of BP neural network to the wood textural samples reached to 98.5%. The result indicates that to recognize wood with the fused features is quite effective.