基于SVM和小波的木材纹理分类算法

Wood texture classification algorithm based on SVM and wavelet

选用二进正交小波基对木材纹理图像进行多层分解,利用所得到的纹理特征向量分析水平、垂直和对角方向上木材纹理频率分布特点。基于木材纹理的这种频率分布特点,选取能够表达木材纹理特征的一组向量作为SVM分类的特征向量,利用多类SVM分类器对木材纹理样本进行训练和识别分类。实验表明,文中基于SVM和小波的木材纹理分类方法优于传统的分类方法。

Based on wavelet method, it realized multi-resolution decompositionof wood surface texture, and analyzedfrequency traits of wood texture at horizontal, vertical and angular directions by eigenvalues from decomposition subsections. Accordingto thesefrequency traits of wood texture, a group of features of wood textureisextracted and used as eigenvalues of multi-class SVM. Examinations indicate that the proposed algorithm obtained competitive results.