Research On Plank Classification Based On Color Texture Features

The color and texture of the surface of the board are the most intuitive characteristics of the surface of the wood,and they are also an important indicator for the quality of wood products in the furniture,construction and decoration industries.This is closely related to the economic benefits of wood processing enterprises.At present,the domestic board processing industry has not formulated a set of industry standards that can systematically evaluate the color and texture of the surface of the board.This topic has studied this issue in depth from the perspective of computer vision.Five tree species,including oak,red pine,white birch,ash and larch,which are common in Northeast China,are selected as the research objects,and the related theoretical methods of image processing and pattern recognition are used to analyze the surface of the above-mentioned tree species in chord and diameter cuts.Parameters,researched out parameter methods that can characterize the surface color and texture characteristics of the board,so as to help realize automatic classification of the board.This article first introduces the basic concept of texture,expounds the definition,type and characteristics of texture,introduces the four main texture feature extraction methods of statistical method,frequency spectrum method,model method and structure method,and compares the above methods.In terms of color analysis of the sheet surface,several common color spaces are introduced,including RGB,CMY,L*a*b*,HSV,etc.,and two main methods for extracting color features are also described.In the part of image classification methods,the principles and design methods of three commonly used classifiers are introduced,namely BP neural network,KNN and support vector machine.This paper adopts a sheet texture feature extraction method based on multi-channel Gabor filtering and Tamura texture features,which overcomes the problem that traditional methods are not sensitive to local texture features when extracting global features of sample images.Specifically,the Tamura texture feature based on visual psychology is combined with the Gabor filter.The texture feature parameter is extracted from the imaginary convolution image of a total of 24 filters in different frequencies and different directions.The texture feature parameters are extracted through the color histogram and color.The two methods of moments extract color features on the sample surface,and combine the above-mentioned texture feature parameters to perform classification experiments on BP neural network,KNN and support vector machine classifiers.The recognition rate of the best feature parameter system is 97.8%.Finally,a support vector machine classification experiment based on particle swarm optimization algorithm was carried out.The classification performance of the optimized classifier improved by 1.3%compared with that before optimization,and the recognition rate reached 99.1%.This topic compares the feature extraction methods of different textures and colors,determines the parameter system and classifier of the color texture features on the surface of the board,and conducts a classification experiment on the board.