彩色自然纹理分类识别的研究

Research on Classification and Recognition in Color Natural Texture

为了建立描述彩色自然纹理的参数体系,选用木材纹理这种典型彩色自然纹理为研究对象,提取了木材图像颜色矩和高斯-马尔可夫特征参数,采用改进的模拟退火算法进行了特征优化筛选及特征级参数融合,得到纹理参数体系并进行分类识别实验,识别率为97.25%,明显高于融合前单参数体系。研究表明用该参数体系对彩色木材纹理进行分类识别是可行的,也可用于与其相近彩色自然纹理的分类识别。

In order to establish the parameters system to describe color natural texture,wood texture is adopted as die classic research object.The features of color moment and GMRF are extracted from wood texture images,and the fused features are obtained by the improved simulated annealing algorithm.With the fused features,the recognition rate reaches to 97.25%.The result indicates that the fused features are feasible to recognize color wood texture,as well as other color nature textures similar to it.