Evaluation of paddy yield and quality estimation methods based on various vegetation indices, NDSI and PLS using BRDF-corrected airborne hyperspectral data

Abstract

This paper describes evaluation of paddy yield and quality estimation methods using an airborne hyperspectral sensor, AISA. Estimation methods are based on various vegetation indices (VIs), Normalized Difference Spectral Index (NDSI), and Partial Least Squares (PLS). In the result of analysis, the paddy quality as measured by the crude protein of brown rice has had a good collection for AISA data, on the other hand the paddy yield haven't. Among vegetation indices, the modified Normalized Difference Vegetation Index (mNDVI) had the highest coefficient of determination (0.61). NDSI combination of about 700 nm and 1600 nm showed the best determination coefficient of 0.70. The determination coefficient using PLS was 0.73. NDSI and PLS using a hyperspectral data appear to be effective for precise estimation of paddy quality.