IMPROVEMENT OF ESTIMATION METHOD FOR GRAMINEOUS CROP PRODUCTIVITY USING NORMALIZATION OF HYPERSPECTRAL DATA

摘要

This paper describes the normalization of hyperspectral data in order to construct a high accuracy estimation method for gramineous crop productivity. Generally, the estimation accuracy of crop is affected by the shadow caused by the topography and the object structure. Some vegetation indices of multispectral sensor are expected to reduce the influence of shadow, therefore previous studies used vegetation indices for estimation of crop productivity. Recently, needs for remote sensing trend with higher accuracy, so that hyperspectral data having high potential estimation capability have been employed in remote sensing applications. However, the technique of reducing shadow influence included in hyperspectral data is still not carefully reviewed. This study suggests the unit vectorized reflectance (UVR), which is one of the normalization for spectral data. The normalization is expected to reduce the shadow influence included in hyperspectral data. Our results show that UVR reduced the effect of shadow and improved the determination coefficient of dry matter pasture from 0.58 to 0.83, as well as the protein content rate of rice from 0.76 to 0.84. While, in comparison with different sun elevation, UVR improves of estimation accuracy under low sun elevation, which increases the influence of shadow. This study shows that the suggested method achieves to reduce shadow influence of hyperspectral data, as well as UVR is effective for improving estimation accuracy for gramineous crop productivity.