基于高光谱和图像技术的苹果叶片叶绿素和磷素含量估测研究

发布时间：2018-08-17 14:27

【摘要】：叶绿素(Chlorophyll)和磷素(Phosphorus)是苹果树生长发育的重要营养元素,是作物健康生长和影响产量的物质基础。传统的叶绿素和磷素测定方法,多为实验室化验分析法,结果虽较为准确,但存在费时、费力的缺点。近年来发展起来的高光谱遥感和图像分析等无损检测技术,因其具有方便快捷的优点,能够实现无损、快速和准确地估测植物营养状况,对提高苹果树的信息化管理具有重要的技术指导与实践意义。本研究以山东烟台栖霞和临沂蒙阴苹果园为研究区,以红富士苹果叶片为研究对象,分别于2014年和2015年5月(新梢旺长期)前后进行样本采集和试验数据测定。利用ASD Field Spec 4地物光谱仪和数码相机分别获取苹果叶片的光谱反射值和图像,在实验室里利用化学分析方法测定叶片叶绿素含量和磷素含量。通过数据分析,得到了苹果叶片磷素含量的原始光谱反射率的响应规律和相关关系,又对原始光谱反射率进行一阶微分变换,得到了一阶微分形式的响应规律和相关关系;构建并筛选了磷素含量的高光谱特征参量,建立了磷素含量的估测模型。在进行图像分割与颜色值的获取基础上,分析了苹果叶片叶绿素含量与RGB颜色参数的相关关系,筛选出了影响叶绿素含量的核心颜色参数,建立了叶绿素含量的估测模型。主要结果有:(1)得到了苹果叶片磷素含量的高光谱敏感波段。经过相关分析,苹果叶片磷素含量与350~2500 nm波段的高光谱反射率的呈显著负相关特性,其中,蓝光(521~568 nm)、红光(697~736 nm)、近红外(1347~1878 nm和2022~2400 nm)波段是磷素含量的敏感波段,其中在R1720处取得最高的相关系数r=-0.6485。(2)筛选出了不同叶绿素含量的苹果叶片核心颜色参数。经过苹果叶片图像直方图分析,构建并筛选了叶片叶绿素含量与RGB颜色系统的核心颜色参数,分别为B值、B/R、B/G、G/(R+G+B)、B/(R+G+B)、(R-B)/(R+B)、(GB)/(G+B)、(R-B)/(R+G+B)、(G-B)/(R+G+B)。(3)建立了苹果叶片磷素含量的估测模型。经比较分析,苹果叶片磷素含量的最佳估测模型为基于主要植被指数的变量组合(DVI(556,712),DVI(677,1728),RVI(542,1094),RVI(705,937),DVI(FDR567,FDR1980)基于高光谱和图像技术的苹果叶片营养状况估测研究NDVI(937,549),和DVI(FDR523,FDR1883))建立的随机森林模型,其估测模型的决定系数R2=0.9236,均方根误差RMSE=0.0158,相对误差为RE=6.9150%。(4)建立了苹果叶片不同叶绿素含量估测模型。基于敏感颜色参数B值、B/R、B/G、G/(R+G+B)、B/(R+G+B)、(R-B)/(R+B)、(G-B)/(G+B)、(RB)/(R+G+B)、(G-B)/(R+G+B)建立的叶绿素含量的支持向量机模型(SVM),对Chl.a、Chl.b、Chl.(a+b)和SPAD估测的决定系数R2分别为0.8754、0.8374、0.8671和0.8129,均方根误差RMSE分别为0.0194、0.0350、0.0497和0.9281,相对误差RE分别为0.8059%、1.7540%、1.122%和1.1894%,模型评估指标均通过了P=0.01极显著性检验水平,尤其是对Chl.a的估测效果最佳。
[Abstract]:Chlorophyll (Chlorophyll) and phosphate (Phosphorus) are important nutrient elements for apple tree growth and development. The traditional methods for the determination of chlorophyll and phosphorus are mostly laboratory analytical methods. Although the results are more accurate, they have the disadvantages of time-consuming and laborious. Non-destructive testing techniques, such as hyperspectral remote sensing and image analysis, have been developed in recent years. Because of their convenient and rapid advantages, they can be used to estimate the nutritional status of plants quickly, quickly and accurately. It has important technical guidance and practical significance for improving the information management of apple trees. In this study, Qixia apple orchard in Yantai, Shandong Province and Mengyin apple orchard in Linyi were used as the research area, and the leaves of Red Fuji apple were used as the research objects. The samples were collected and the experimental data were measured before and after May 2014 and May 2015. The spectral reflectance and image of apple leaves were obtained by ASD Field Spec 4 ground object spectrometer and digital camera respectively. The chlorophyll content and phosphorus content in leaves were determined by chemical analysis method in laboratory. Through the data analysis, the response law and the correlation relation of the original spectral reflectance of apple leaf phosphorus content were obtained, and the first order differential transformation of the original spectral reflectance was carried out, and the response law and the correlation relation of the first order differential form were obtained. The hyperspectral characteristic parameters of phosphorus content were established and the estimation model of phosphorus content was established. On the basis of image segmentation and color value acquisition, the correlation between chlorophyll content and RGB color parameters of apple leaves was analyzed, the core color parameters affecting chlorophyll content were screened out, and the estimation model of chlorophyll content was established. The main results were as follows: (1) the hyperspectral sensitive band of phosphorus content in apple leaves was obtained. The correlation analysis showed that the phosphorus content in apple leaves was negatively correlated with the hyperspectral reflectance at 350 ~ 2500nm. The blue (521 ~ 568 nm),) red (697 ~ 736 nm),) near infrared (1347 ~ 1878 nm and 2022 ~ 2 400 nm) bands were sensitive to phosphorus content. The highest correlation coefficient was obtained at R1720. (2) the core color parameters of apple leaves with different chlorophyll content were selected. Based on histogram analysis of apple leaf image, the core color parameters of leaf chlorophyll content and RGB color system were constructed and screened. The estimation model of P content in apple leaves was established as B value B / R / (R G B) B / (R G B), (R B) / (R B), (GB) / (G B), (R-B / (R G B), (G-B) / (R G B). (3). By comparative analysis, the optimal estimation model of phosphorus content in apple leaves was established as a random forest model based on the variable combination of main vegetation index (DVI (556712) and RVI (572 / 1094) / RVI (705937) DVI (FDR567N / FDR1980), NDVI (937549) and DVI (FDR523UFDR1883) based on hyperspectral and image techniques. The determination coefficient of the estimation model is R2 + 0.9236, the root mean square error (RMSE) is 0. 0158, and the relative error is RET 6. 9150. (4) the estimation model of different chlorophyll content in apple leaves has been established. A support vector machine model based on the sensitive color parameter B / B / (R G B) / (R G B), (G-B / (G B), (RB) / (R G B), (G-B) / (R G B) was established to estimate the determination coefficients of Chl. (a b) and SPAD for Chl. Chl. (a b) and SPAD were 0.87540.83740.8671 and 0.8129, respectively. The root mean square error (RMSE) was 0.01943.500.0497 and 0.9281, respectively.The mean square error (RMSE) was 0.01943500.0497 and 0.9281respectively, and the determination coefficients of (a b) and SPAD were 0.87540.83740.8671 and 0.8129, respectively. The mean square error (RMS) was 0.01943500.0497 and 0.9281respectively. For the error RE of 0.80590.75400.122% and 1.1894%, the model evaluation index has passed the P0.01 significant test level. In particular, the estimation of Chl.a is the best.
【学位授予单位】：山东农业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S661.1

【参考文献】