数学
查阅表格
天蓬
叶面积指数
遥感
大气辐射传输码
偏最小二乘回归
辐射传输
统计
计算机科学
地理
农学
光学
物理
生物
考古
程序设计语言
作者
Shanqin Wang,Weizhen Gao,Ming Jin,Lantao Li,Dihong Xu,Shishi Liu,Jianwei Lu
标识
DOI:10.1016/j.compag.2018.07.023
摘要
Inversion of radiative transfer models (RTM) provides an avenue for assessment of crop status in precision agriculture. The potential of PROSAIL inverted using Tree-structure Parzen Estimators (TPE), a hyper-parameter searching algorithm to retrieve crop variables was evaluated in this study using a simulated dataset and an actual field experiment dataset. For simulated dataset, the carotenoid content (Car), brown material (Cb) and equivalent water thickness (Cw) were estimated with high accuracies using the simulated leaf area index (LAI), leaf dry mass (LMA) and canopy chlorophyll content (Cab) as input variables. Even using LAI as the only input variable in the PROSAIL model, LMA, Cab and Car could be estimated with reasonable accuracies. The performances of Partial Least Squares Regression (PLSR) and Lookup Table (LUT) based PROSAIL inversion were tested likewise on the field dataset. Compared with PLSR and LUT-based approaches, the TPE-based approach estimated Cab with the highest accuracy (R2 = 0.82, nRMSE = 0.10) using LAI and LMA as known canopy variables to invert PROSAIL with maximum 10,000 iterations set in TPE. Cab were estimated respectively using PLSR and LUT-based approaches with reasonable accuracy, while LMA was only estimated using LUT-based approach when the LUT entries were created with known LAI. Our results reveal that the TPE-based inversion of the PROSAIL model is a promising method to retrieve canopy variables with inputs of canopy reflectance and nondestructively measured variables.
科研通智能强力驱动
Strongly Powered by AbleSci AI