The Preliminary Research On Hyperspectral Prediction Model Of Leaf Nitrogen Content, Production And Relevant Quality Of Rice Population

The test was conducted in the experimental farm of Yangzhou University and soil culture pool during2012-2013, which using Liangyoupeijiu and Wuyunjing24as material and planting on four nitrogen level (0,150,225,300kg/hm2). Analysing the canopy spectral difference between two varieties in jointing stage, heading stage and mature stage, and aiming to seek the best growth period to estimate rice nitrogen, yield and quality, the study respectively built hyperspectral remote sensing models of two varieties to estimate leaf nitrogen content, yield and quality, which has tested the accuracy. The results showed that:(1) In different growth period, the canopy spectral of two rice varieties under different nitrogen treatment had the same characteristics, but the trend of reflection spectrum curve is consistent. The reflection peak formed at550nm, and the reflectance decreased with the increase of nitrogen application. The near infrared reflection platform is formed after750nm, and the reflectance increased with the increase of nitrogen application. In different growth period, the canopy spectral reflectance of the same variety is different, the reflectance of heading stage is maximum and the mature stage is minimum. The canopy spectral reflectance between different varieties is different, the canopy spectral reflectance of Wuyunjing24was higher than liangyoupeijiu.(2) The hyperspectral optimal model of leaf nitrogen content of two varieties in different growth period is different. For liangyoupeijiu, the regression equation of jointing stage which based on canopy spectral parameter Db is y=340406x2-3297.3x+9.9575, RMSE is0.2189, RE is17.87%; the regression equation of heading stage which based on spectral parameter SDr/SDb is y=0.043x2-0.2543x+1.3419, RMSE is0.0441, RE is1.22%; the regression equation of mature stage which based on spectral parameter SDr is y=267.96x2-233.77x+ 53.123; RMSE is0.0712, RE is6.58%. For Wuyunjing24, the regression equation of jointing stage which based on canopy spectral parameter (SDr-SDb)/(SDr+SDb) is y=-135.11x2+195.85x-67.791, RMSE is0.1524, RE is7.35%; the regression equation of heading stage which based on spectral parameter SDr/SDb is y=-0.0009x2+0.1188x+0.6046, RMSE is0.0657, RE is1.89%; the regression equation of mature stage which based on spectral parameter SDr is y=528.53x2-458.74x+101.79, RMSE is0.4567, RE is7.87%.(3) The hyperspectral optimal model of grain yield of two varieties were respectively y=-100061x2+240567x-143908based on spectral parameter (SDr-SDy)/(SDr+SDy) for liangyoupeijiu, which RMSE is0.0842, RE is1.11%, and y=-522.63x2+3145.5x-4063based on spectral parameter SR705for Wuyunjing24, which RMSE is0.1271, RE is1.83%.(4) The hyperspectral optimal model of grain amylose content of two varieties were respectively y=-0.4884x2+3.0388x+17.187based on spectral parameter SDr/SDb for liangyoupeijiu, which RMSE is0.1870, RE is1.27%, and y=0.0892x2-2.8479x+33.027based on spectral parameter SDr/SDb for Wuyunjing24, RMSE is0.1045, RE is1.13%. The hyperspectral optimal model of grain protein content were respectively y=-23.5x2+56.962x12.928based on spectral parameter ND705for liangyoupeijiu, RMSE is0.3236, RE is2.18%, and y=-968.48x+1716.6x-751.34based on spectral parameter NDVI for Wuyunjing24, RMSE is0.1321, RE is1.17%.(5) The hyperspectral model of nitrogen content, production and grain amylose content and protein content which related to quality of rice is the best in heading stage.