224-1 Estimation of Leaf Dry Weight Usinghyperspectral Vegetation Indices and Continuous Wavelet Analysis for Wheat Canopies.
Poster Number 141
See more from this Division: ASA Section: Agronomic Production SystemsSee more from this Session: General Precision Agriculture: II
Tuesday, November 4, 2014
Long Beach Convention Center, Exhibit Hall ABC
LDW (green leaf dry weight per ground area) is the key variable for describing crop growth. Real-time, dynamic and accurate monitoring and diagnosis of LDW are important for evaluating crop growth and improving crop nutrient management. This study was based on eight field experiments that obtained systematic information on canopy spectral reflectance and LDW of wheat plants over a 7-years period (10 varieties at 5 sites). Models for monitoring LDW in wheat leaves were constructed using the vegetation index (VI) and continuous wavelet analysis approaches (CWA).
The results showed that among the different spectral indices, DSI (R1688, R922) was the best for monitoring LDW in wheat (R2 = 0.730, RMSE = 0.0266 kg/m2, RRMSE = 0.224). A systematic study using 108 wavelet functions from 15 wavelet families showed db3, db4, db6, db7, db8, db16, db17, db20, sym3, and gaus3 performed better; whereas, shan1-1, fbsp1-1-1.5, fbsp2-1-1, cmor1-1, cmor-1.5, cgau4, cgau5, cgau6, cgau7, and cgau8 performed slightly worse. The db7 (1197 nm, scale 8) was the best for monitoring LDW (R2 = 0.747, RMSE = 0.032 kg/m2, RRMSE = 0.272). Both VI and CWA can be used to estimate LDW in wheat in the NIR region (780 ~ 1,350 nm) with better performance for CWA than that for VIs. The results demonstrate that the newly developed LDW model on the method of VI and CWA could improve the performance of LDW estimation under different ecological site sand varied nitrogen treatments, while providing the theoretical basis of monitoring crop growth by hyperspectrum.
See more from this Division: ASA Section: Agronomic Production SystemsThe results showed that among the different spectral indices, DSI (R1688, R922) was the best for monitoring LDW in wheat (R2 = 0.730, RMSE = 0.0266 kg/m2, RRMSE = 0.224). A systematic study using 108 wavelet functions from 15 wavelet families showed db3, db4, db6, db7, db8, db16, db17, db20, sym3, and gaus3 performed better; whereas, shan1-1, fbsp1-1-1.5, fbsp2-1-1, cmor1-1, cmor-1.5, cgau4, cgau5, cgau6, cgau7, and cgau8 performed slightly worse. The db7 (1197 nm, scale 8) was the best for monitoring LDW (R2 = 0.747, RMSE = 0.032 kg/m2, RRMSE = 0.272). Both VI and CWA can be used to estimate LDW in wheat in the NIR region (780 ~ 1,350 nm) with better performance for CWA than that for VIs. The results demonstrate that the newly developed LDW model on the method of VI and CWA could improve the performance of LDW estimation under different ecological site sand varied nitrogen treatments, while providing the theoretical basis of monitoring crop growth by hyperspectrum.
Key words: Wheat; Hyperspectrum; LDW; Vegetation index; Continuous wavelet transform; Monitor model
See more from this Session: General Precision Agriculture: II
Previous Abstract
|
Next Abstract >>