基于Gabor滤波与Tamura纹理特征的板材分类研究

板材表面的纹理特征是木质板材表面最为直观的特性，同时也是建筑装潢质量和木制品品质的重要评价指标。以中国东北部常见的红松、落叶松、白桦、水曲柳和柞木等5种树种的弦切、径切图像作为研究对象，提出一种基于多通道Gabor滤波和Tamura纹理特征的板材纹理特征提取方法，克服了传统方法在提取样本图像的全局特征时对局部纹理特征不敏感的问题。具体是将基于视觉心理学的Tamura纹理特征与Gabor滤波器进行结合，在不同频率、不同方向上共24组滤波器的虚部卷积图像上进行纹理特征参数提取，结合上述的纹理特征参数在BP神经网络、KNN和支持向量机分类器上进行分类试验，最佳特征参数体系的识别率达97.8%。

Plank Classification Based on Gabor Filtering and Tamura Texture Features

The texture feature of the surface of a board is the most intuitive characteristics of the wooden board,and it is also an important evaluation index for the quality of building decoration and wood products.In this paper,the string cutting and diameter cutting images of five commonly occurring tree species in northeastern China were taken as the research objects,including Pinus koraiensis,Larix gmelinii,Betula platyphylla,Fraxinus mandshurica,and Xylosma racemosum.A method of extracting the plate texture features based on multi-channel Gabor filter and Tamura texture features was proposed,which overcame the problem of insensitive to local texture features when extracting the global features of sample images used in traditional method.Specifically,the Tamura texture feature based on visual psychology was combined with the Gabor filter,and the texture feature parameters were extracted on the imaginary convolution image of a total of 24 filters in different frequencies and different directions,combined with the above texture feature parameters,the classification experiment was carried out on the BP neural network,KNN and support vector machine classifier,and the recognition rate of the best feature parameter system reached 97.8%.