联合主成分分析与最小二乘支持向量机估测冬小麦叶面积指数

利用单一植被指数估测叶面积指数存在高光谱遥感丰富的波段信息易丢失和外界因素干扰大的缺点,但若将波段信息全部引入模型又会增加建模难度。为解决利用多波段信息估测叶面积指数的问题,利用主成分分析法(PCA)对光谱数据进行降维,之后将提取的主成分与最小二乘支持向量机(LS-SVM)模型相结合,构建冬小麦叶面积指数的高光谱估测模型,并与以4类植被指数作为LS-SVM输入参数建立的模型进行比较。结果表明,以主成分作为LS-SVM模型的输入参数建立的模型精度最高,模型检验集R2为0.71,检验集RMSE为0.56,估测结果较使用植被指数作为输入参数建立的模型精度高,稳定性好。该方法可为利用多波段信息进行大范围冬小麦叶面积指数的无损测定提供参考。

Estimation of Winter Wheat Leaf Area Index with Joint Principal Component Analysis and Least Squares Support Vector Model

The method of inverting leaf area index using a single vegetation index can lead to loss of rich hyperspectral band information , and the modeling will be difficult if all band information introduced into model. In order to solve the problem of LAI estimation using the multi band information is a question. In order to solve this problem, the dimensionality of spectral data was reduced by using principal component analysis (PCA), then the extracted information was used with least squares support vector machine (LS SVM) regression algorithm to establish the estimation model of winter wheat LAI, and the model was compared with the 4 models based on single vegetation index as a input of LS SVM. Results showed the accuracy of LS SVM model was the highest, with test set decision coefficient R of 0.71, RMSE of inversion value and measured value was 0.56. The estimated results were superior to single vegetation index on precision and stability, which could provide reference to estimate winter wheat LAI in a wide range by using multi band information.