高光谱遥感的冬小麦籽粒蛋白质含量监测

发布时间：2018-05-13 12:17

本文选题：冬小麦 + 蛋白质含量　；参考：《安徽理工大学》2017年硕士论文

【摘要】：随着科学的进步和经济的发展,农产品产量问题得到重视的同时,品质问题也越来越受关注。籽粒蛋白质含量(Grain Protein Content,GPC)是作物品质性状之一,本研究针对GPC遥感监测这一科学问题从两方面展开分析:一方面,考虑到蛋白质的形成是一个日积月累的形成过程,从简单经验模型出发,建立基于多时相光谱参量的GPC估算模型。另一方面,从籽粒氮素运转原理出发,结合籽粒氮素累积量(Grain Nitrogen Accumulation,GNA)与籽粒产量(Grain Yield,GY)的比值最终实现GPC的估算。得到的结论主要有:(1)采用灰色关联分析和偏最小二乘算法,发现多生育期的作物光谱估测籽粒蛋白质含量精度更高。后三期(挑旗期、开花期、灌浆期)特征累积值构建的GPC模型具有较高的可靠性,与单一生育期模型精度最高的时期(开花期)相比,估算决定系数明显提高,均方根误差明显减小。(2)利用典型植被指数累积值进行籽粒产量的研究。发现,开花期的归一化植被指数(744,784)与冬小麦产量的相关性最大;非线性模型精度高于线性模型,并且后三期预测模型精度与单生育期模型精度最高的时期(开花期)相比,决定系数明显提高,均方根误差明显降低。因此,多生育期累积值反演精度高于单生育期,说明多生育期归一化植被指数累积值预测籽粒产量具有一定的可行性。(3)利用籽粒氮素运转原理和波段深度分析方法对籽粒氮素累积量进行研究。GNA由开花前植株贮存向籽粒运转的氮(Nitrogen Translocation Amount,NTA)和花后植株同化的氮(Nitrogen Assimilation Amount,NAA)组成。通过植株氮素运转原理,将NTA的反演转化为叶片氮素累积量的反演。经计算发现,利用植被指数FD742反演开花期叶片氮素累积量最为理想,最终形成植被指数反演NTA的模型。另外,利用波段深度分析和偏最小二乘,分析开花期和灌浆期特征指数与NAA的相关性。发现,特征指数在两个生育期与NAA决定系数均大于0.7,达到极显著相关水平。(4)GNA与GY的比值通过换算系数得到籽粒蛋白质含量模型。通过耦合模型进行估算籽粒蛋白质含量,综合考虑相关系数和均方根误差的大小发现,灌浆期的特征指数归一化面积波段深度反演NAA进而反演GPC的精度最高,且建模和验证的预测值及真实值之间变化保持一致。
[Abstract]:With the development of science and economy, more and more attention has been paid to the quality of agricultural products. Grain protein content (Protein) is one of the traits of crop quality. In this study, the scientific problem of GPC remote sensing monitoring was analyzed from two aspects: on the one hand, considering that protein formation is a cumulative process, Based on the simple empirical model, the GPC estimation model based on multiphase spectral parameters is established. On the other hand, based on the principle of grain nitrogen transport, the ratio of grain Nitrogen accession (GNA) and grain yield (GY) was combined with the ratio of grain nitrogen accumulation and grain yield to realize the estimation of GPC. The main conclusions are as follows: (1) using grey correlation analysis and partial least squares algorithm, it is found that the precision of estimating grain protein content by crop spectrum in multi-growth stage is higher. The GPC model constructed by the accumulative values of the latter three periods (flag picking, flowering, filling) was of high reliability, and the estimated determinant coefficient was significantly higher than that of the single growth stage model with the highest precision (flowering period). The root mean square error (RMS) was significantly reduced. (2) the grain yield was studied by using the accumulation value of typical vegetation index. It was found that the normalized vegetation index (NDVI) of flowering stage (744784) had the highest correlation with winter wheat yield, and the accuracy of nonlinear model was higher than that of linear model, and the precision of prediction model of the latter three periods was higher than that of single growth stage model (flowering period). The coefficient of determination is increased and the root mean square error is obviously reduced. Therefore, the accuracy of inversion of cumulative value in multiple growth period is higher than that in single growth period. The results show that it is feasible to predict grain yield by the accumulation value of normalized vegetation index in multi-growth stage.) the grain nitrogen accumulation is studied by using the principle of nitrogen transport and the depth analysis method of band. GNA is stored by the plant before flowering. Nitrogen Translocation Amountna (NTAA) and N Nitrogen Assimilation AmountNAA (assimilated after anthesis). Based on the principle of plant nitrogen transport, the inversion of NTA was transformed into the inversion of leaf nitrogen accumulation. It is found by calculation that the best way to invert nitrogen accumulation in leaves at flowering stage is to use vegetation index (FD742) to invert nitrogen accumulation, and finally to form a model of vegetation index inversion for NTA. In addition, the correlation between the characteristic index of flowering and filling stage and NAA was analyzed by using band depth analysis and partial least square method. It was found that the coefficient of determination between characteristic index and NAA was greater than 0.7 at two growth stages, and the ratio of GNA-GY to GY reached the most significant correlation level. The model of grain protein content was obtained by conversion coefficient. The coupling model was used to estimate the grain protein content. Considering the correlation coefficient and the root mean square error, it was found that the characteristic index normalized area band depth inversion of NAA and the inversion of GPC at filling stage had the highest accuracy. And the change between the predicted value and the real value of modeling and verification is consistent.
【学位授予单位】：安徽理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S512.11;S127

【相似文献】