小麦隐蔽性害虫检测新模型  

A new model of detecting hidden insect in wheat

作　　者：史卫亚[1] 乔娜娜[1] 梁义涛[1] 王锋[1] 

机构地区：[1]河南工业大学信息科学与工程学院//粮食信息处理与控制教育部重点实验室,河南郑州450001

出　　处：《粮食与饲料工业》2014年第10期66-70,共5页Cereal & Feed Industry

基　　金：国家863计划(2012AA101608);国家自然科学基金项目(31171775);河南教育厅自然科学项目(14B550002)

摘　　要：粮食在存储过程中极易发生虫蚀现象,因此需要快速高效的检测粮食是否染虫。基于生物光子分析技术(Biophoton Analytical Technology,BPAT),提出了一种新型的小麦隐蔽性害虫检测模型。以小麦籽粒和玉米象为研究对象,分别测量正常和含虫小麦的自发光子数,并提取8个统计特征(均值、方差、中位数、四分位数、平均差、离散系数、偏度、峰度)和13个直方图特征组成粮食特征数据向量,对这些特征向量进行主成分分析(principal component analysis,PCA),并在此基础上分别采用线性分类器LDA(linear discriminant analysis)和BP神经网络模型进行识别,实验结果表明,所提模型可以较好地区分正常小麦和含虫小麦。Insect damage is easy to occure in grain storing, t.herefore a fast and efficient method is needed to detect the potential risk. Based on the biophoton analytical technology （BPAT）, a new model of detecting hidden insect was proposed. Taking wheat kernel and maize weevil as research objects, spontaneous photon numbers of uninfested and infested wheat were measured representatively. Grain characteristic data vectors were consisted of eight statistical characteristics （mean, median, variance, quartile, mean difference, discrete coefficient, skewness and kurtosis） and 13 histogram features. Then the data was anlyzed by the Principal Component Analysis （PCA）. Based on it, the feature vectors were tested using linear discriminant analysis （LDA） and BP neural network model for identification. Results showed that the proposed method was able to well differentiate uninfested and infested wheat.

关 键 词：小麦 玉米象 虫蚀 生物光子 统计特征 直方图特征 识别 

分 类 号：S435.122[农业科学—农业昆虫与害虫防治]