| 基于SVM的加工番茄早疫病叶氮素含量光谱反演 |
| |
| 引用本文： | 尹小君,张 清,赵庆展,汪传建,宁 川.基于SVM的加工番茄早疫病叶氮素含量光谱反演[J].农业机械学报,2014,45(11):280-285. |
| |
| 作者姓名： | 尹小君 张 清 赵庆展 汪传建 宁 川 |
| |
| 作者单位： | 石河子大学;中国科学院遥感与数字地球研究所数字地球重点实验室;石河子大学;石河子大学;石河子大学 |
| |
| 基金项目： | “十二五”国家科技支撑计划资助项目（2012BAH27B02）、国家自然科学基金资助项目（31260291）、中国科学院数字地球重点实验室开放基金资助项目（2012LDE011）和石河子大学高层次人才基金资助项目（RCZX201226） |
| |
| 摘 要： | 采用支持向量机(SVM)模型,对新疆加工番茄早疫病病害植株的叶片氮素含量进行光谱反演。分析不同病害严重度的病叶氮素含量的光谱特征,发现在218~357 nm、384~587 nm、1 033~1 141 nm、1 499~2 500 nm,氮素含量与光谱反射率的相关系数的绝对值大于0.7,在227~353 nm的相关系数大于0.8,表明不同病害严重度的病叶氮素含量与光谱反射率呈强相关。利用K层交叉检验(K-CV)方法验证、优选出SR705、ND705、GMI-2、RI-half、PTEBc等5种光谱指数,作为SVM模型的输入变量;同时,分别建立线性核、多项式核、径向基核和Sigmoid核的SVM模型,通过模型拟合比较,得出最佳模型为径向基核的SVM模型。采用径向基核的SVM模型对病叶氮素含量进行光谱反演,结果表明:径向基核的SVM模型氮素含量反演的真实值与预测值的MSE为0.012 4,相关系数R为85.916%,平均相对误差为0.175,结合多光谱指数的SVM模型提高了加工番茄早疫病病害叶片氮素含量的反演精度。
|
| 关 键 词： | 加工番茄 早疫病 氮素含量 光谱指数 反演 支持向量机 |
| 收稿时间： | 2013/11/19 0:00:00 |
Remote Sensing Inversion of Nitrogen Content Based on SVM in Processing Tomato Early Blight Leaves |
| |
| Yin Xiaojun,Zhang Qing,Zhao Qingzhan,Wang Chuanjian and Ning Chuan.Remote Sensing Inversion of Nitrogen Content Based on SVM in Processing Tomato Early Blight Leaves[J].Transactions of the Chinese Society of Agricultural Machinery,2014,45(11):280-285. |
| |
| Authors: | Yin Xiaojun Zhang Qing Zhao Qingzhan Wang Chuanjian and Ning Chuan |
| |
| Institution: | Shihezi University;Key Laboratory of Digital Globe, Institute of Remote Sensing and Digital Earth of CAS;Shihezi University;Shihezi University;Shihezi University |
| |
| Abstract: | Support vector machine was used to invert nitrogen content of processing tomato early blight leaves in Xinjiang. The spectrum characteristic of processing tomato of difference disease level was analyzed. Then nitrogen content was found to be strong correlation with the spectral reflectivity on 218~357nm, 384~587nm, 1033~1141nm,1499~2500nm, because the correlation coefficients were more than 0.8. The vegetation index, SR705, ND705, GMI-2, RI-half, and PTEBc were chosen through K-CV cross validation, and SVM model was used to invert the nitrogen content with the vegetation index. The results show that the precision the SVM model of radial basis function kernel was the highest in linear kernel, polynomial kernel, radial basis function kernel and Sigmoid kernel. The value of MSE was 0.0124. The value of R was 85.916%. The value of average relative error was 0.175. SVM model with multi-vegetation index improved the precision of inverting nitrogen content of processing tomato early blight leaves. |
| |
| Keywords: | Processing tomato Early blight Nitrogen content Spectrum index Inversion Support vector machine |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《农业机械学报》浏览原始摘要信息 |
| 点击此处可从《农业机械学报》下载免费的PDF全文 |
|
