欢迎访问《中国农学通报》,

中国农学通报 ›› 2014, Vol. 30 ›› Issue (7): 286-291.doi: 10.11924/j.issn.1000-6850.2013-2289

• 植物保护 农药 • 上一篇    下一篇

基于支持向量机的昆虫数值化鉴定

吴宏华 张红燕 陈渊   

  • 收稿日期:2013-08-28 修回日期:2013-09-11 出版日期:2014-03-05 发布日期:2014-03-05
  • 基金资助:
    国家自然科学基金青年基金项目“基于基因协同作用的肿瘤特征基因选择及其共表达网络分析”(61300130);国家科技支撑计划重大项目“农村物联网基础平台共性关键技术研究”(2012BAD35B05);湖南省研究生科研创新项目“时间序列分析方法在农业虫害预测中的应用研究”(CX2012B307)。

Digital Insect Identification Based On Support Vector Machine

  • Received:2013-08-28 Revised:2013-09-11 Online:2014-03-05 Published:2014-03-05

摘要: 为提高昆虫鉴定的准确度,基于支持向量机提出了一种新的计算机昆虫数值化鉴定方法,并应用于以前翅内部翅脉交点距离为数值特征的7种蝴蝶的鉴定。首先利用DrawWing软件对7种蝴蝶的翅脉交点坐标进行了自动获取,并计算各相邻交点之间的欧式距离;然后将每类样本与其他样本组成二分类模型;再对每一模型经支持向量机非线性特征筛选,去除无用或冗余特征值,并以保留特征构建最终分类器。7个预测模型的独立测试平均精度达98.64%,明显高于参比模型,表明新方法在昆虫鉴定领域具有较好的应用前景。

关键词: 检测方法, 检测方法

Abstract: Based on support vector machine (SVM), a novel method for digital insect identification was proposed, and applied in identifying seven species of butterflies with intersectional coordinates of venation in the internal of forewings. The basic principles were as follows: firstly, the intersectional coordinates of venations in the internal of seven species’ forewings were obtained automatically by DrawWing which was a program for numerical description of insect wings. Secondly, binary model was composed by the each type of sample and other samples. Thirdly, the redundant features or unnecessary features were filtered by using support vector classification, and the retained features were used to construct the classification model. Accuracy of the seven prediction model was 98.64%, and higher than the reference model, that the new method of identification in the field of insect has a good prospect.