橡胶树割胶期氮素水平的叶片光谱检测机理及诊断模型研究

As the main method of nutrient detection of rubber tree, biochemical assay of rubber leaves has a long dull procedure and can not satisfying the practical requirement.So,a quickly method of rubber tree's nitrogen content dagnosis is hotly demanded in farm manage. But, there are a lot of work to do for the aim. This research will focus on the rubber tree's nitrogen content dagnosis by spectroscopy.Starting from the spectrum of rubber leaves,the work will cover the correlation between original spectrals, sensitive spectral band,spectral feature parameter and qualitative & quantitative diagnosis.The research is aimed to find a mechanism of spectrum detection of rubber leaves's nitrogen ,to porpose a model of nitrogen content of rubber tree,and to provide theoretical and technical for rubber tree planting. There are three parts which are contained in the study. Firstly, our research will investigate the correlation of leaves's raw spectrum between rubber seedlings and rubber tree, and reveal the sensitive bands of leave's raw spectrum which got from rubber tree about nitrogen content. Secondly, the topic will study the selection method of spectral characteristics which based on compression algorithm of spectral data and statistical algorithm of spectral general characteristics, discuss the multi-scale and consistency of spectral data,and then obtain the spectral feature parameters of nitorgen dagnosis in rubber tree. Finally, the research will construct the qualitative & quantitative model for nitrogen content dagnosis, propose the steps and algorithm of dagnosis which is mainly based on the models. The content of algorithm evaluation is the last part.

由于传统生化分析技术繁琐低效，橡胶树田间管理对氮素快速检测需求迫切，但目前对割胶期橡胶树氮素水平的快速检测技术研究还很不够。本项目以橡胶树割胶期氮元素的检测为主线，以橡胶树叶片的光谱响应研究为切入点，采用"原始光谱相关性-光谱敏感波段-光谱特征参数提取-定性定量综合诊断"的技术路径，研究橡胶树叶片氮素水平的光谱检测机理，提出橡胶树割胶期氮素水平的诊断模型，为提高橡胶树割胶期的现代管理水平提供理论和使能技术支撑。研究包括：（1）研究橡胶种苗和割胶期橡胶树叶片原始光谱的相关性，明确割胶期橡胶树叶片氮素光谱检测敏感波段。（2）研究基于光谱数据总体特征、组合优选等波段压缩算法的光谱特征选择方法，研究光谱数据的多尺度分析和一致性信息求解的算法，提取割胶期橡胶树叶片氮素营养的光谱特征参数。（3）研究叶片氮素营养光谱检测的定性判识和定量诊断的模型，研究"定性-定量"递进式光谱诊断的方法，完成模型的评价。

本项目研究了割胶期橡胶树叶片氮素含量光谱诊断的基础理论与方法，主要成果包括：首先，考虑到氮素光谱敏感波段是揭示橡胶树叶片光谱响应机理的重要信息，详细分析了橡胶树叶片氮素原始光谱数据的相关性，并基于分数阶微积分、间隔随机蛙、连续投影算法等提出了相应的敏感波段筛选算法，基本确定割胶期橡胶树叶片的氮素敏感波段在可见光谱区间为500nm~525nm和700nm~750nm，在近红外光谱~间为1000~1150nm、1760~1810nm以及2000~2200nm段。其次，为了提取可以由于构建诊断模型的橡胶树叶片氮素的光谱特征向量，利用分数阶微积分、直接正交信号处理等预处理手段，并在相关性分析的基础上，提出了基于多模谱线敏感特征的融合特征向量、基于相关系数的连续投影算法的低共线性特征向量，从多个角度对光谱数据信息进行数据描述。最后，面向不同的应用需求，从定性和定量两个角度研究了割胶期橡胶树叶片氮素光谱诊断模型。提出并构建了基于BP-Adaboost、极限学习机、偏最小二乘等算法的定性、定量诊断模型。通过本项目的实施，对橡胶树叶片氮素含量的快速诊断问题中涉及的相关关键科学问题进行了分析，为后续相关应用设备的研发提供了新思路和理论依据。