| 水稻害虫图像识别技术研究 |
| |
| 引用本文： | 李文斌.水稻害虫图像识别技术研究[J].安徽农业科学,2014(23):8043-8045. |
| |
| 作者姓名： | 李文斌 |
| |
| 作者单位： | 杭州电子科技大学通信工程学院,浙江杭州310018 |
| |
| 摘 要： | 目的]解决水稻害虫传统识别方法的低时效性问题。方法]采用数字图像处理方法对水稻害虫进行图像识别和分类,对水稻害虫的虫体面积、虫体周长、偏心率、形状参数、似圆度、叶状性、球形性等几何形状特征进行提取和研究,并采用支持向量机(SVM)分类器对水稻害虫二化螟、三化螟、稻飞虱、卷叶螟进行分类。结果]利用所建立的6个特征判别函数对4种水稻害虫进行判别分类,识别率达到96.67%,说明这6个经过筛选的特征具有很强的判别性。结论]支持向量机分类器的识别方法很好地解决水稻害虫传统识别方法的低时效性问题。支持向量机以风险最小化为原则,兼顾训练误差与测试误差的最小化,具体体现在分类模型的选择和模型参数的选择上。
|
| 关 键 词： | 图像处理 特征提取 识别分类 |
Research for Rice Pests Image Recognition Technology |
| |
| LI Wen-bin.Research for Rice Pests Image Recognition Technology[J].Journal of Anhui Agricultural Sciences,2014(23):8043-8045. |
| |
| Authors: | LI Wen-bin |
| |
| Institution: | LI Wen-bing (College of Communication Engineering, Hangzhou University of Electronic Science and Technology, Hangzhou, Zhejiang 310018) |
| |
| Abstract: | Objeetive ] The aim was tu solve the low timeliness issue of the traditional identification methods. Method] Digital image pro- cessing method was used to identity and class the rice pests in]ages, and to extract and research their geometrical eunfiguratiml features, such as pests area, perimeter, eecentrieity, shape parameters, roundness and the other geometric shape thatures, then support vector machine （SVM） classifier was used to class rice pests, such as Chilo suppressalis Walker, Tryporyza ineerlulas Walker, Nilaparvata lugens Sial. Le-rodea euihla Edwards. Resuh ] The recognition rate of 4 kinds of pests was high and reached 96.67% by using established 6 kinds of charac- teristic critical functions, showing this screened 6 kinds of characteristics had strong disc.rim｜nation. Conclusion ] The recognition method based on SVM better solves the low timeliness issue of the traditional identification methods. SVM is based on the prineiple of risk minimiza- film, considers both training error and test error, it is embodied by both selection of elassification model and selection of model parameters. |
| |
| Keywords: | huage process Feature extraction Identification and classification |
| 本文献已被 CNKI 维普 等数据库收录！ |
|
