油菜的高光谱特征及其生理参数估算模型研究

发布时间：2018-06-26 00:37

本文选题：高光谱遥感 + 油菜　；参考：《西北农林科技大学》2017年硕士论文

【摘要】：农业是一个国家的基本产业,属于第一产业,是提供支撑国民经济建设与发展的基础产业。油菜的平均产量在国产植物油中占比超过了一半。同时,油菜的产油效率较高,因此在我国植物油产业中具有重要地位。因此快速,准确,大范围,实时地获取油菜相关的生理参数,检测油菜长势情况显得尤为重要。本研究以油菜为研究对象,通过田间试验和实验室检测,分别获取了油菜冠层高光谱遥感数据,不同尺度的油菜高光谱遥感影像以及油菜的生理参数。通过计算机和数理统计方法,构建油菜最佳的生理参数预测模型,并通过不同尺度的高光谱遥感图像反演验证,为油菜的精细农业发展奠定基础,推动了农业数字信息化进程。取得的主要结论如下:(1)从整体趋势上来看,不同生育期油菜冠层高光谱特征较为一致,在可见光蓝紫光和红光波段范围有两个吸收谷,在绿光范围有明显的反射峰,在近红外光范围形成极高反射区域。同时随着冠层叶绿素(SPAD)含量以及氮素含量的增加,在可见光范围内油菜冠层高光谱反射率降低,在近红外光范围内的极高反射区域升高。随着冠层叶绿素(SPAD)含量以及含氮量的增加,红边特征也出现了“红移”的现象。从苗期到开花期,红边特征的“红移”现象明显,等到了成熟期,红边特征逐步显现出“蓝移”的现象。由此可以看出,油菜冠层的生理参数与高光谱反射率之间存在这密不可分的联系。(2)通过相关处理油菜冠层光谱数据,提取了多个能够反映植被生理参数的光谱特征参数和植被指数。同时分别提取与这些光谱参数对应的不同生育期油菜的叶绿素(SPAD)含量和叶片氮素含量。然后利用这些提取的数据进行相关性分析,结果证明绝大多数光谱特征参数和植被指数都与叶绿素(SPAD)含量和叶片氮素含量都是极显著相关,同时全生育期油菜构建的叶绿素(SPAD)含量和叶片氮素含量预测模型不仅精度高于分生育期油菜,还避免了单生育期预测模型无法预测其他生育期油菜生理参数的弊端。(3)在对全生育期油菜的叶绿素(SPAD)含量和叶片氮素含量预测研究过程中发现,通过随机森林回归算法,并选择自变量影响力较好的光谱特征参数和植被指数作为构建模型的自变量,构建的预测模型精度好于普通回归算法和多元逐步回归算法。在前人研究的基础上,提高了油菜的相关生理参数预测模型的精度。(4)利用随机森林回归算法构建的油菜的叶绿素(SPAD)含量和叶片氮素含量预测模型,对不同尺度的油菜高光谱影像进行反演分析,结果发现反演结果基本符合油菜真实养分分布情况。
[Abstract]:Agriculture is the basic industry of a country, belongs to the primary industry, and provides the basic industry to support the construction and development of the national economy. The average yield of rapeseed accounted for more than half of the domestic vegetable oil. At the same time, rapeseed has high oil production efficiency, so it plays an important role in Chinese vegetable oil industry. Therefore, it is very important to obtain the physiological parameters of rapeseed quickly, accurately, in large range and in real time, and to detect the growth of rapeseed. In this study, the rape canopy hyperspectral remote sensing data, the hyperspectral remote sensing images of rape on different scales and the physiological parameters of rape were obtained by field experiments and laboratory tests. Through computer and mathematical statistics method, the best physiological parameter prediction model of rapeseed was constructed, and the inversion verification of hyperspectral remote sensing image of different scales was carried out, which laid a foundation for the development of fine agriculture of rape and promoted the process of agricultural digital information. The main conclusions are as follows: (1) from the overall trend, the hyperspectral characteristics of rape canopy at different growth stages are consistent, and there are two absorption valleys in the band of visible blue violet light and red light, and obvious reflection peaks in the green light range. An extremely high reflectance region is formed in the near infrared region. At the same time, with the increase of chlorophyll (Spad) content and nitrogen content in the canopy, the hyperspectral reflectance of rape canopy decreased in the visible light range, and the extremely high reflectance region increased in the near infrared region. With the increase of chlorophyll (Spad) content and nitrogen content in the canopy, "red shift" appeared in the red edge. From seedling stage to flowering stage, the "red shift" phenomenon of red edge characteristics was obvious, and the "blue shift" phenomenon appeared gradually at maturity stage. It can be seen that there is a close relationship between the physiological parameters of rape canopy and hyperspectral reflectance. (2) processing rape canopy spectral data by correlation, Several spectral characteristic parameters and vegetation indices which can reflect the physiological parameters of vegetation were extracted. At the same time, the chlorophyll (Spad) content and leaf nitrogen content of rape at different growth stages were extracted respectively. The results showed that most of the spectral characteristic parameters and vegetation index were significantly correlated with chlorophyll (Spad) content and leaf nitrogen content. At the same time, the prediction precision of chlorophyll (Spad) content and leaf nitrogen content of rapeseed at the whole growth stage was higher than that of the rapeseed at different growth stage. It also avoids the disadvantage that the single growth stage prediction model can not predict the physiological parameters of other growth stages of rape. (3) in the process of prediction of chlorophyll (Spad) content and leaf nitrogen content in rapeseed during the whole growth period, it was found that the stochastic forest regression algorithm was used. The spectral characteristic parameters and vegetation index which have better influence on independent variables are chosen as independent variables to construct the model, and the prediction model is better than ordinary regression algorithm and multivariate stepwise regression algorithm. On the basis of previous studies, the precision of the prediction model of relative physiological parameters of rapeseed was improved. (4) the prediction model of chlorophyll (Spad) content and leaf nitrogen content of rapeseed was constructed by using stochastic forest regression algorithm. The inversion analysis of different scales of rape hyperspectral images shows that the inversion results basically accord with the true nutrient distribution of rapeseed.
【学位授予单位】：西北农林科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S565.4

【参考文献】