Abstract:
The aim of this work was divided into 2 main parts. The first part was to compare the potential of hyperspectral and multispectral remote sensing coupled with Field Spectroradiometer to discriminate the 3 varieties of rice which were RD41, RD 31 and Pathum Thani 1. The second part was to separate the RD-41 rice with and without bacterial leaf blight disease at Nang Lue-Tha Chai sub-district, Muang district, Chainat province, Thailand. The retrieving spectral reflectance from Field Spectroradiometer and a scene of the EO-1 Hyperion imagery by the Genetic Algorithm based band selector coupled with Maximum Likelihood Classifier (MLC) and Spectral Angle Mapper (SAM) techniques in order to reduce a data set with a large number of dimensions. From the results obtained, the overall accuracy of the classification of rice RD 41, RD43 and Pathum Thani 1 undergo 3 growth phases from hyperspectral remote sensing with the SAM provided higher overall accuracy (OA) and Kappa statistics than GA coupled with MLC of both field level and satellite level. The overall accuracy from satellite level using EO-1 Hyperion showed more powerful classifier than EO-1 ALI sensor. For bacterial leaf blight disease investigation, the overall accuracy of the SAM classifier presented the value of OA and Kappa statistics from satellite level using EO-1 Hyperion which was higher than the value of EO-1 ALI sensor OA and Kappa statistics. The overall accuracy obtained of both rice classification and bacterial leaf blight disease separation from hyperspectral remote sensing technique was significantly higher than Multispectral remote sensing (p-value < 0.05). The usefulness of these results should be offered a guideline for a variety of rice classification and rice leaf disease detection.