论文部分内容阅读
以ASD FieldSpec-Vnir光谱仪实测不同生长季大豆的冠层反射率,同期采集对应大豆LAI,然后逐波段分析冠层光谱反射率、导数光谱与大豆LAI的相关关系;并采用单变量线性回归逐波段分析了冠层光谱反射率、导数光谱与大豆LAI确定性系数随波长的变化趋势,建立了以近红外与可见光波段冠层光谱反射率的比值植被指数RVI与大豆LAI的高光谱遥感估算模型。结果表明,冠层光谱反射率在350￣680nm、760￣1050nm波谱区与大豆LAI相关性较大,而在红边区680￣760nm的相关性变化较大;导数光谱在红边区与大豆LAI相关程度高。通RVI方式建立的遥感估算模型能较为准确估算大豆LAI,通过对红外与蓝波段建立的RVI指数与大豆LAI的回归模型,表明其预测大豆LAI的能力较好,有进一步研究的必要;通过对比发现,神经网络模型可以大大提升高光谱反演大豆LAI的水平,模型的确定系数R2为0.9661,而总均方根误差RMSE仅为0.446m2.m-2。
The ASD FieldSpec-Vnir spectrometer was used to measure the canopy reflectance of soybean in different growing seasons. The corresponding LAI of soybean was collected at the same period. Then the correlation between canopy spectral reflectance and soybean LAI was analyzed by band-by-band. Single- The spectral reflectance, derivative spectrum and determinant coefficient of soybean LAI with wavelength were analyzed. Hyperspectral remote sensing estimation model of ratio vegetation index (RVI) and soybean LAI with near-infrared and visible-light spectral reflectance was established. The results showed that the spectral reflectance of the canopy was in the range of 350-680 nm, the correlation of the spectral range of 760-1050 nm with soybean LAI was relatively large, while the correlation of the spectral range of 680-760 nm was greater in the red border area. high. The remote sensing estimation model established by the RVI method can accurately estimate the LAI of soybeans. The regression model of the RVI index and soybean LAI established in the infrared and blue bands indicates that the LAI predicting ability of soybeans is better, which is necessary for further research. It was found that the neural network model can significantly improve the LAI of hyperspectral inversion, the model determination coefficient R2 is 0.9661, while the total root mean square error RMSE is only 0.446m2.m-2.