摘要
【目的】以福建省顺昌县大干镇的毛竹为研究对象,研究毛竹叶片氮元素含量的最优估测模型,为毛竹生长状态分析与林地土壤肥力估测提供基础。【方法】通过对毛竹叶片原始光谱、一阶微分光谱及相关的植被指数与叶片氮元素含量进行相关性分析来筛选氮元素敏感特征参数,并构建了多元线性回归模型、随机森林模型以及支持向量机模型,利用决定系数最优原则筛选3个模型中的最优模型并进行精度验证。【结果】R387、DR663、NDVIg-b(R575、R440)、SIPI、PRI和PPR 6个参数与毛竹叶片氮含量具有较为显著的相关性,基于这6个敏感参数所构建的3种模型中,多元线性回归模型与随机森林模型拟合效果较差,精度验证结果R2分别为0.4355、0.4371,惩罚因子C和核参数Sigma分别设为3和0.1的支持向量机模型估测结果最好,其实测值与预测值拟合决定系数为0.8031,总体精度为94.02%。【结论】基于R387、DR663、NDVIg-b(R575、R440)、SIPI、PRI和PPR 6个叶片光谱参数所构建的支持向量机模型能够较为准确地估测毛竹叶片氮元素含量。
【Objective】Taking the bamboo from Dagan town,Shunchang county,Fujian province as the research object,the optimal estimation model of nitrogen content in bamboo leaves was studied,which provided the basis for the analysis of bamboo growth status and forest soil fertility estimation.【Methods】The sensitivity parameters of nitrogen elements were screened by correlation analysis between original spectra,first-order differential spectra,relevant vegetation indices and nitrogen content of leaves,and multiple linear regression models,random forest models and support vectors were constructed.The optimal model of the three models was screened by comparing the coefficient values,and then the accuracy was verified.【Result】The six parameters of R387,DR663,NDVIg-b(R575,R440),SIPI,PRI,and PPR have a significant correlation with the nitrogen content of bamboo leaves.Among the three models,the multiple linear regression model and the random forest model have a poor fitting effect.The coefficient of determination of the accuracy verification result is 0.4355,0.4371 respectively.The Support vector machine model with the penalty factor C and the kernel parameter Sigma set to 3 and 0.1 is the best of the three models with a coefficient of determination of 0.8031 and an overall accuracy of 94.02%.【Conclusion】The support vector machine model constructed by using six spectral parameters of R387,DR663,NDVIg-b(R575,R440),SIPI,PRI and PPR can accurately estimate the nitrogen content of bamboo leaves.
作者
林灵辰
余坤勇
曾琪
姚雄
邓洋波
范华栋
刘健
LIN Lingchen;YU Kunyong;ZENG Qi;YAO Xiong;DENG Yangbo;FAN Huadong;LIU Jian(College of Forestry,Fujian Agriculture and Forestry University,Fuzhou 350002,Fujian,China;3S Technology and Resources Optimized Utilization Key Laboratory of Fujian University,Fuzhou 350002,Fujian,China)
出处
《中南林业科技大学学报》
CAS
CSCD
北大核心
2020年第2期81-87,130,共8页
Journal of Central South University of Forestry & Technology
基金
“十三五”国家重点研发子课题(2018YFD060010304)
福建省高校产学研重点项目(2015N5010)
关键词
毛竹
氮元素
叶片光谱
随机森林
支持向量机
moso bamboo
nitrogen
leaf spectrum
random forest
support vector machine
