欢迎访问《麦类作物学报》编辑部网站

作者	单位
何汝艳，蒋金豹，郭海强，郭会敏，陶亮亮	（1.中国矿业大学（北京）地球科学与测绘工程学院，北京 100083 2.北京师范大学环境演变与自然灾害教育部重点实验室，北京 100875）

摘要点击次数: 1199

全文下载次数: 1184

中文摘要:

为综合更多有效信息来提高冬小麦叶绿素密度的估测精度，应用投影寻踪降维方法对条锈病胁迫下冬小麦冠层光谱进行降维，生成一维向量，然后采用支持向量机回归方法对其叶绿素密度进行估测，并与高光谱植被指数估测结果进行了比较。结果表明，以小麦冠层一阶微分光谱与叶绿素密度相关性较高的波段（400~500 nm、720~770 nm和840~870 nm）进行投影寻踪降维得到的最优一维向量为自变量，利用支持向量机回归方法构建的冠层叶绿素密度估测模型的精度最高，决定系数为0.867，均方根误差与相对误差均最小，分别为1.135 μg·cm^-2和13.6%。说明利用投影寻踪降维技术对条锈病胁迫下冬小麦冠层光谱进行降维处理，可以保留有效信息，提高冬小麦叶绿素密度估测精度。

英文摘要:

In this paper, the projection pursuit (PP) dimension reduction was applied to process the canopy reflectance data of winter wheat under the stress of the stripe rust to obtain an optimal vector being as the independent variable, then the support vector machine (SVM) regression method was used to build the model of estimating the canopy chlorophyll density (CCD) of winter wheat. Several hyperspectral vegetation indices were also chosen as the independent variable to estimate the CCD of winter wheat by SVM regression method, and the results of the above two methods were compared. The results showed that the model based on PP dimension reduction of first derivative canopy spectra, which acquired an optimal vector being as the independent variable in 400~500 nm, 720~770 nm and 840~870 nm, had the highest estimation precision than other models, the R² was 0.867, the RMSE was 1.135 μg·cm^-2 and relative error was 13.6%. Above result suggested that the PP dimension reduction method could reduce the dimensions of canopy spectra data of winter wheat, and could retain the useful and valid spectral information as much as possible, thus it could improve the precision of the model for estimation the CCD of winter wheat under the disease stress. Therefore, the PP dimension reduction method could provide important information for monitoring the crop growth status and diseases using hyperspectral remote sensing.

查看全文查看/发表评论下载PDF阅读器

关闭