Abstract
It is well known that the protein content of rice grain is an indicator of taste of cooked rice in the countries where people as the staple food. Ground-based optical sensing over the crop canopy would provide information not only on the mass of plant body which reflects the light, but also on the crop nitrogen content which is closely related to the greenness of plant leaves. The vegetation index has been related to crop variables such as biomass, leaf nitrogen, plant cover, and chlorophyll in cereals. The objective of this study was to investigate the correlation between GNDVI and NDVI values, and grain protein content at different dates and to estimate the grain protein content using G(NDVI) values. We measured Green normalized difference vegetation index [$GNDVI=({\rho}0.80{\mu}m-{\rho}0.55{\mu}m)/({\rho}0.80{\mu}m+{\rho}0.55{\mu}m)$] and [$GNDVI=({\rho}0.80{\mu}m-{\rho}0.68{\mu}m)/({\rho}0.80{\mu}m+{\rho}0.68{\mu}m)$] by using two different active sensors. The study was conducted during the rice growing season for three years from 2005 through 2007 at the experimental plots of National Institute of Agricultural Science and Technology. The experiments were carried out by randomized complete block design with the application of four levels of nitrogen fertilizers(0, 70, 100, 130kg N/ha) and the same amount of phosphorous and potassium content of the fertilizers. After heading stage, relationships between GNDVI of rice canopy and grain protein content showed the highly positive correlation at different dates for three years. GNDVI values showed higher correlation coefficients than that of NDVI during growing season in 2005-07. The correlation between GNDVI values at different dates and grain protein contents was highly correlated at early July. We attempted to estimate the grain protein content at harvesting stage using GNDVI values from early July for three years. The determination coefficients of the linear model by GNDVI values were 0.9l and the measured and estimated grain protein content at harvesting stage using GNDVI values highly correlated($R^2=0.96^{***}$). Results from this study show that GNDVI appeared very effective to estimate leaf nitrogen and grain protein content of rice canopy.
본 연구에서는 광학 센서를 이용한 벼 생육단계 별 식생지수와 쌀 단백질함량의 관계를 구명하여 수확기 쌀 단백질함량을 추정하고자 하였다. 인공광원을 사용하는 능동형 광학센서인 GreenSeeker(NTech Inc., USA) GNDVI(green normalized difference vegetation index=$({\rho}0.80{\mu}m-{\rho}0.55{\mu}m)/({\rho}0.80{\mu}m+{\rho}0.55{\mu}m)$)와 NDVI(normalized difference vegetation index=$({\rho}0.80{\mu}m-{\rho}0.68{\mu}m)/({\rho}0.80{\mu}m+{\rho}0.68{\mu}m)$) 2종의 센서를 이용하여 벼 군락의 반사특성을 측정하고 동시에 식물체 샘플링을 통한 쌀 단백질함량을 분석하였다. 3년 동안(2005-2007년) 벼 출수 후 식생지수와 쌀 단백질함량의 관계를 조사해 본 결과 모든 시기에 걸쳐 GNDVI가 NDVI보다 상관이 높았고. 벼 수확기가 가까울수록 상관계수가 높게 나타났다. 수확기 쌀 단백질함량 예측 가능성을 알아보기 위해 벼 유수형성기와 출수기 두 시기의 GNDVI값과 수확기 쌀 단백질함량과의 관계를 분석해본 결과, 결정계수가 각각 0.91, 0.81로 특히 이삭거름 주기 전에 측정한 GNDVI를 통하여 수확기 쌀 단백질함량을 예측 할 수 있다는 결론을 얻었다. 이 결과를 바탕으로 유수형성기 GNDVI를 이용한 수확기 쌀 단백질함량 경험 모델식을 구하고 경험 모델식에서 얻어진 추정값과 실측값의 관계를 통해 검증하였다. 2005년과 2006년에서 구한 경험모델식의 쌀 단백질함량 추정값과 2007년도 쌀 단백질함량 실측값을 1:1 line에서 비교해본결과 결정계수가 높게 나타났다($R^2=0.96^{***}$).
