基于Total多分类支持向量机的小麦产地判别分析  

Discriminatory Analysis for the Growing Area of Wheat Based on Total Multi-Class Support Vector Machine

作　　者：徐云[1] 吴静珠[2] 石庆兰[1] 冯磊[1] 徐义田[3] 

机构地区：[1]中国农业大学信息与电气工程学院,北京100083 [2]北京工商大学计算机与信息工程学院,北京100037 [3]中国农业大学理学院,北京100083

出　　处：《沈阳农业大学学报》2013年第3期333-336,共4页Journal of Shenyang Agricultural University

基　　金：国家重点实验室开放项目(2011-SKL-0);中央高校基本科研业务项目(2012QJ088);国家自然科学基金项目(61153003;10771213;60977057)

摘　　要：采用近红外光谱分析的方法快速鉴别小麦的产地,可为小麦流通监管、优质小麦生产基地保护和国际贸易提供理论和技术支持。以来自中国不同地域的202份小麦样品为例,在已知样品组分含量(蛋白质、湿面筋、沉降值、硬度)的前提下,结合样品的近红外光谱信息,利用Total多分类支持向量机对小麦产地进行判别分析。在构建分类模型的过程中,使正确分类的小麦样品尽可能地远离分类超平面,使错误分类的小麦样品尽可能地靠近分类超平面,可获得较高的分类精度。通过100个小麦样品构建分类模型,对另外的102个小麦样品产地进行预测,有80个与实际产地相符,预测精度为78.43%。为小麦产地鉴别提供一种新的方法。Rapid identification of wheat producing area through near infrared spectrum analysis provides theoretical and technical support for circulation regulation, high-quality wheat production base protection and international trade. In this paper, 202 wheat samples come from different regions in China and their sample compositions （protein, wet gluten, sedimentation value, hardness） are used for wheat origin discrimination analysis, based on near infrared spectrmn of sample and Total classification support vector machine （SVM）. Moreover, it not only makes the correctly classified samples as far as possible from the classification hyper-plane, but also makes the incorrectly classified samples as near as possible to the hyper-plane. Then it yields high prediction accuracy. We construct the classification hyper-plane by 100 training samples, and then to predict the other 102 testing samples. There are 80 samples corresponding with their practical growing area, then the prediction is 78.43%, which tests the validity of algorithm.

关 键 词：Total多分类支持向量机 近红外光谱 小麦产地 判别分析 

分 类 号：S511[农业科学—作物学]