摘要
提出了一种有效描述白酒显微图像中微粒的形状和结构分布的组合特征提取方法,研究了基于显微图像的白酒识别方法.显微图像先经过全差分滤波增强,以及互熵阀值区域分割.然后提取显微图像的形状和结构特征,包括所有区域面积、周长和低阶矩的总和、均值及方差,最大5个区域的畸度和圆度,以及区域数目,共计26个特征值.最后基于BP神经网络建立白酒识别模型.通过实验,比较了传统形状特征与文中提出的形状及结构组合特征在白酒识别中的效果,结果表明使用文中提供的形状及结构组合特征获得更高的识别率,达到95%以上.
A new combinational feature extraction method is introduced in the paper,which can efficiently describe the structure and region shape of wines' particles in micrograph,and the classification of Chinese wines based on the micrograph is also studied.First,the micrographs are enhanced using total variation filter,and segmented using relative entropy threshold.Then 26 features of the shape and structure are extracted using proposed method in the paper,including the sum,average value and variance of the regions' area,perimeter and moments,as well as the eccentricity and circularity of the biggest five regions,and the total number of all regions.Finally,Chinese wine classification system based on micrograph using combination of shape and structure features and BP neural network have been presented.The authors compare the recognition results for different choices of features(traditional shape features or proposed features).The experimental results show that the better classification rate has been achieved using the combinational features proposed in this paper,and the classification rate is over 95%.
出处
《四川师范大学学报(自然科学版)》
CAS
CSCD
北大核心
2011年第4期593-596,共4页
Journal of Sichuan Normal University(Natural Science)
基金
四川省教育厅自然科学重点基金(2006A166)
教育厅重大培育项目基金(07ZZ017)资助项目
关键词
显微图像
白酒鉴别
结构特征
形状特征
BP神经网络
micrograph
classification of wine
structure feature
shape feature
BP neural network