Predicting TN67 (Oryza Sativa) yield using canopy reflectance after different pretreatments
| Autor: | Chang-Li Liu, 劉長利 |
|---|---|
| Rok vydání: | 2007 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 95 The reflectance spectra from the remote sensing images can be used to predict the yield of the rice. However, the reflectance spectra data should be pre-processed to solve the problem of colinearity and noise. Some pretreatment methods, e.g. standard normal variate (SNV), multiplicative signal correction (MSC), and orthogonal signal correction (OSC) were employed to correct the noise before establishing the model. In order to keep off the colinearity between the independent variable, we may replace with the suitable independent variable or use the multivariate statistical analysis method. In the study, blue band (BLUE), green band (GRN), red band (RED), near infrared band (NIR), normalized difference vegetation index (NDVI), green normalized difference vegetation index (GNDVI), sample ratio vegetation index (SRVI), and green ratio vegetation index (GRVI) were used to build the first order linear models. In addition, the reflectance spectra data (between 350nm to 1100nm by 10nm) were employed to build the PLSR models after using SNV, MSC, and OSC pretreatment methods. Because of the small sample size, the cross validation was implemented to estimate the prediction ability. From this study, it was found that the best predicted ability of the rice yield model was the PLSR model after using the OSC pretreatment. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|