基于主成分分析的矿物浮选泡沫图像分类与识别  

CLASSIFICATION AND RECOGNITION OF MINERAL FLOTATION FROTH IMAGES BASED ON PRINCIPAL COMPONENT ANALYSIS

作　　者：王红平[1] 齐春[1] 李金标[2] 张忠信[3] 

机构地区：[1]西安交通大学电子与信息工程学院,西安710049 [2]昆明冶金研究院,昆明650031 [3]云南会泽铅锌矿,云南曲靖654211

出　　处：《矿冶》2005年第3期79-82,共4页Mining And Metallurgy

摘　　要：浮选泡沫图像是一种特殊的纹理图像,不同类别等级泡沫图像比较相似,采用邻域灰度相关矩阵法或空间灰度相关矩阵法提取特征参数,然后用神经网络进行分类的效果不是很理想。本文利用主成分分析法对特征参数进行变换处理,改变不同类别特征参数的聚集程度,然后利用神经网络进行分类。对比实验结果表明:对特征参数进行特征变换处理后,分类的正确率大大提高。Flotation froth images belong to a special kind of texture images. But its classification is not very ideal through using neural network analysis of the feature parameters obtained by neighboring gray level dependence matrix approach and other similar approaches owing to the different rank of froth images is so similar. Better classification could be reached with the use of principal component analysis approach to transform the original feature parameters into new feature parameters favorable to BP neural network for the implementation of the classification of flotation froth images. Comparison tests showed that the classification based on the transformed feature parameters is much accurate than that on the original feature parameters.

关 键 词：浮选 泡沫图像 主成分分析 BP神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TD91[自动化与计算机技术—计算机科学与技术]