Fu Yuan-Yuan, Wang Ji-Hua, Yang Gui-Jun, Song Xiao-Yu, Xu Xin-Gang, Feng Hai-Kuan
Institute of Applied Remote Sensing & Information Technology, Zhejiang University, Hangzhou 310029, China.
Guang Pu Xue Yu Guang Pu Fen Xi. 2013 May;33(5):1315-9.
The major limitation of using existing vegetation indices for crop biomass estimation is that it approaches a saturation level asymptotically for a certain range of biomass. In order to resolve this problem, band depth analysis and partial least square regression (PLSR) were combined to establish winter wheat biomass estimation model in the present study. The models based on the combination of band depth analysis and PLSR were compared with the models based on common vegetation indexes from the point of view of estimation accuracy, subsequently. Band depth analysis was conducted in the visible spectral domain (550-750 nm). Band depth, band depth ratio (BDR), normalized band depth index, and band depth normalized to area were utilized to represent band depth information. Among the calibrated estimation models, the models based on the combination of band depth analysis and PLSR reached higher accuracy than those based on the vegetation indices. Among them, the combination of BDR and PLSR got the highest accuracy (R2 = 0.792, RMSE = 0.164 kg x m(-2)). The results indicated that the combination of band depth analysis and PLSR could well overcome the saturation problem and improve the biomass estimation accuracy when winter wheat biomass is large.
利用现有植被指数估算作物生物量的主要局限性在于，对于一定范围的生物量，它会渐近地接近饱和水平。为了解决这个问题，本研究将波段深度分析和偏最小二乘回归（PLSR）相结合，建立冬小麦生物量估算模型。随后，从估算精度的角度，将基于波段深度分析和PLSR相结合的模型与基于常见植被指数的模型进行了比较。波段深度分析在可见光谱域（550 - 750 nm）进行。利用波段深度、波段深度比（BDR）、归一化波段深度指数以及面积归一化波段深度来表示波段深度信息。在校准的估算模型中，基于波段深度分析和PLSR相结合的模型比基于植被指数的模型具有更高的精度。其中，BDR与PLSR相结合的模型精度最高（R2 = 0.792，RMSE = 0.164 kg x m(-2)）。结果表明，波段深度分析与PLSR相结合能够很好地克服饱和问题，并在冬小麦生物量大时提高生物量估算精度。
