基于小波分析的大豆叶面积高光谱反演

Wavelet transformation of in-situ measured hyperspectral data in Glycine max LAI estimation

实测了不同水肥耦合、经营制度及有效营养面积条件下的大豆(Glycinemax)冠层高光谱反射率与叶面积指数(LAI),并对光谱反射率、微分光谱与LAI的关系进行了分析;采用比值植被指数(RVI)与归一化植被指数(NDVI)建立了大豆LAI反演模型;采用小波分析对采集的光谱反射率数据进行了能量系数提取,并以小波能量系数作为自变量进行了单变量与多变量回归分析,对大豆LAI进行估算。结果表明:大豆LAI与光谱反射率在可见光波段呈负相关;在近红外波段呈正相关;微分光谱在红边处与大豆LAI密切相关(R2=0.92);RVI与NDVI可以提高大豆LAI的估算精度(R2分别达0.79、0.84);各植被指数各有优缺点,应根据需要进行选择;小波能量系数回归模型可以进一步提高大豆叶面积的估算水平,以一个特定小波能量系数作为自变量的回归模型,大豆LAI回归确定系数R2高达0.884;以4个和6个小波能量系数建立LAI回归分析模型(R2分别达0.92、0.93),2个模型LAI预测值与大豆LAI实测值线性回归确定性系数R2分别为0.90、0.92。比较可知,小波分析可以对高光谱进行特征变量提取,进而反演大豆生理参数,并且反演的LAI精度较光谱反射率、微分光谱及植被指数都有明显提高,小波分析在植被生理参数的高光谱提取方面有着广阔的应用前景。

The canopy reflectance and leaf area index （LAI） of Glycine max under different water-fertilizer coupling and management conditions were measured, and the relations of reflectance and derivative reflectance with LAI were analyzed. Ratio Vegetation Index （RVI） and Normalized Difference Vegetation Index （NDVI） were regressed against LAI, and the wavelet energy coefficients of spectral reflectance were extracted to estimate the LAI. It was found that the canopy reflectance and LAI had a negative relation in visible region but a positive relation in near infra- red region. Reflectance derivative had close relations with LAI in blue, green and red edge spectral regions, with the maximum correlation coefficient in red edge region. RVI and NDVI had close relations with LAI, the regression determination coefficient R^2 being 0.79 and 0.84, respectively. These two vegetation indices had their own advantages and disadvantages, and the selection of them depended on the requirements. A regression model established with single wavelet energy coefficient was obtained, and the determination coefficient R^2 was 0. 884. Step wise regression with 4 and 6 wavelet energy coefficients were also made, and the results showed that the determination coefficient R^2 between the regression models with 4 and 6 independents and the predicted and measured LAI were 0. 90 and 0.92, respectively. It could be concluded that wavelet transform could be applied to in situ collected hyperspectral data processing and estimate G. max LAI much more accurately, in comparing with reflectance, derivative or vegetation indices. Wavelet transform could also be applied to hyperspectral data for the estimation of other vegetation＇ s biophysical and biochemical parameters.