基于数据挖掘算法的竹块颜色分级研究  

Color classification of bamboo pieces based on data mining

作　　者：张昭[1] 罗文亮[1] 乔永亮[2] 何东健[2] 

机构地区：[1]宝鸡文理学院电子电气工程系 [2]西北农林科技大学机械与电子工程学院

出　　处：《中国农机化学报》2013年第2期217-220,共4页Journal of Chinese Agricultural Mechanization

基　　金：国家自然科学基金资助项目(60975007);宝鸡文理学院重点项目(ZK10157)

摘　　要：为了实现计算机视觉对竹块自动识别与分级,研究基于数据挖掘算法的竹块颜色分类方法。该方法先将原始图像由RGB转换为HIS空间并结合OTSU法完成背景分割,用radon变换进行倾斜校正后,提取竹块颜色判定有效区域并计算其HIS 3分量的均值和标准偏差,对获取的颜色特征数据采用C4.5算法构建了分级模型。分级实验结果表明,提出的方法准确率最高达到93.33%,高于BP算法及NB tree算法,为竹块颜色分级方法提供一种有效、可行的方法。We propose a detection system based on color classification and data mining algorithms.Firstly,RGB color is converted to HIS color.Then bamboo object is segmented with threshold estimated by Otsu algorithm.The tilt registration is carried out based on Radon transform.Then the effective region for color classification is extracted and the mean and standard deviation of the pixels in the region for H,S and I component respectively are calculated as the color feature parameters.To obtain the color characteristics of data classification C4.5 decision tree algorithm,build a classification model.The classification results show that the classification model accuracy rate of this model is 95%,significantly higher than that BP and NB tree classification algorithm.Therefore bamboo classification method proposed in this paper,is feasible and effective.

关 键 词：竹块 颜色 分类 神经网络 计算机视觉 

分 类 号：TP206[自动化与计算机技术—检测技术与自动化装置]