Abstract: The floor hierarchical classification based on the color characteristics is proposed to improve the automate degree. Through clustering the pre-specified color characteristics of the floor sample, the broad categories and specific categories are created to determine the training samples. To classify the unknown floor samples, the first use of the shortest distance classification is to determine the general category of the sample, then uses the improved K-nearest neighbor classification to determine the specific categories. Test results show that the hierarchical processing program reduces the K-nearest neighbor of the data processing phase with a higher classification accuracy.

Key words: color moments, floor classification, maximum minimum distance algorithm, minimum distance classification, K-nearest-neighbor classification

摘要： 为提高地板生产过程中分类处理的自动化程度及其快速性，提出了一种基于地板颜色特征的层次分类方法。在HSV色彩空间提取出地板的颜色矩特征，并给予色调特征数据以较高权重，降低纹理特征对分类的影响。在对未知样本分类时，利用预先建立的粗细两层分类依据库，采用最短距离决策、K-最近相邻对未知地板样本由粗到细进行逐层判定。测试结果表明，该分层处理方案在保证较高的分类正确率（95.6%）的基础上，有效地减少了K-最近相邻的数据处理量。

关键词: 颜色矩, 地板分类, 最大最小距离算法, 最短距离分类, K-最近相邻分类