The Research Of Pattern Recognition Method On Wood Surfacer Texture

Wood surface color and texture is an important characteristic of wood surface feature parameters,which reflects the visual of wood surface and psychological feeling,and directly related to the quality,grade,price of wood products and the economic benefits of wood processing enterprises.Research on wood surface color texture classification recognition not only could solve the need of wood processing enterprises in production,but also enrich the field of image processing and pattern recognition for the studies of natural texture classification.The research carries on wood surface texture classification and recognition using image processing and pattern recognition theory,and the main content was as follows.Choose coniferous forest(Pinus-koraiensis,Larix-gmelinii),broad-leaved forest(Betula-platyphylla,Fraxinus-mandshurica,Quercus- mongolica) as research objects,which are common in northeast,and establish the sample database including 1,000(100×10) samples of 10 color texture categories.Different color spaces have different characteristics,the research needed to carry on wood surface texture classification and recognition,which hoped the sample owing a good separability in color space.Therefore,characteristic parameters of the color samples in five spaces were obtainned, which is RGB color space,HSV color space,L^* a^* b^* color space,I₁I₂I₃ color space, Normalization color space,and then the classification experimental based on classification rate combined with color space of their own characteristics was carded on.Finally,when color characteristics of wood surface color texture image were obtained,HSV color space was chosen.Under the HSV color space,10 samples categories of color histogram and color moments characteristic parameters were obtained,and 10 samples categories of the co-occurrence matrix characteristic parameters were obtained under the gray space,HSV color space,RGB color space. Combined the GNFS methods,five wood surface texture parameters systems was founded,which wood surface texture parameter systems were as follows:①(Gray co-occurrence matrix(GLCM) and color histogram was fused to found the texture parameters system ),②(GLCM and color moment was fused to found the texture parameters system),③(GLCM,color histogram and color moment was fused to found the texture parameters system ),④(the texture parameters system was based on RGB color space Co-occurrence Matrix)⑤(the texture parameters system was based on HSV color space Co-occurrence Matrix).The classifiers using in this reseacher include the nearest neighbor(1-NN) classifier,probabilistic neural network(PNN) classifier and the integrated BP neural network classifier. Through the three classifiers using the five wood surface texture parameters of the recognition rate and the dimensions of parameter systems,finally the optimal wood surface texture parameters system was acquired,which was the seconded feature parameters system,[W1,W4,W5,W7, W8,W10,W11,H_m1,S_m1,S_m2,S_m3,V_m1,V_m2,V_m3].The recognition rate of the seconded feature parameter system to unknown wood samples was highly up to 97.00%,and the very satisfied result was gained.