| Author | Poonsak Miphokasap |
| Call Number | AIT Diss. no.RS-12-01 |
| Subject(s) | Sugarcane--Remote sensing
|
| Note | A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Remote Sensing and Geographic Information Systems |
| Publisher | Asian Institute of Technology |
| Series Statement | Dissertation ; no. RS-12-01 |
| Abstract | Sugarcane is one of the most important economic cro ps in Thailand which is used to produce sugar and to generate power. Nutrient defic iencies and intensity of foliar are among the most important factors affecting sugarcan e growth and productivity. Monitoring of crop status in sugarcane is, therefore, essentia l for directly assessing the consequences on yield or indirectly evaluating the adverse plant symptoms which relate susceptibility to pests and diseases. The complex spatial-temporal ch aracteristics of spectral characteristics deriving from the fields might not be explained by linear model. In this situation, non- linear relationship usually gives more flexibility than in simple linear model and returns the better estimation results. This study aimed to investigate the potential of hyperspectral data in estimating Canopy Nitrogen Concentration (C NC) and Leaf Area Index (LAI) of sugarcane at full canopy cover and map the spatial distribution of sugarcane CNC at the space level using linear and non-linear model. The results showed that sugarcane CNC is more accur ately estimated by the new integrated approach, involving Kernel Principal Component Anal ysis (KPCA), continuum-removed absorption features as well as a Support Vector Reg ression (SVR). The highlight of this study is that SVR technique was applied at the firs t time for estimating crop nutrient from hyperspectral data. It can be concluded that the mo del accuracy for estimating CNC can be significantly improved even its cultivars were mixe d, with relative error of 3.04%, 3.5% and 4.07% by narrow vegetation indices, SMLR and SV R, respectively when comparing with the previous publication. Sensitive spectral i nformation, contained in the visible (400- 700 nm), red edge (670-780 nm), and far near-infrar ed (1100-1286 nm) regions of the electromagnetic spectrum was reported in this study . It can be summarized that canopy architecture influences directly to the sugarcane s pectral signature and the predictive model capability. Canopy architecture closely relates to amount of light intensity penetrating the sugarcane canopy and interacting with the subsequen t leaves. The results showed a strong interaction between genetic cultivars and CNC, LAI in effecting spectral reflectance. This provided a basis for the methodologies to use in ma pping sugarcane foliar nutrient in a mixed cultivar environment. The integrating of KPCA and non-linear transformation function based on SVR could explain 78 % of the var iation in sugarcane nitrogen concentration using orbiting hyperspectral data. Th is result was better as compared to the use of narrow vegetation index and multiple linear regression. Some selected spectral wavelengths at the space level differ from the repo rted wavelengths at the field experimental level especially in the middle-infrare d regions of the electromagnetic spectrum. |
| Year | 2012 |
| Corresponding Series Added Entry | Asian Institute of Technology. Dissertation ; no. RS-12-01 |
| Type | Dissertation |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Remote Sensing (RS) |
| Chairperson(s) | Honda, Kiyoshi |
| Examination Committee(s) | Nagai, Masahiko ;Souris, Marc |
| Scholarship Donor(s) | Royal Thai Government ;Asian Institute of Technology |
| Degree | Thesis (Ph.D.) - Asian Institute of Technology, 2012 |