水氮互作下银杏叶绿素含量与颜色特征的相关分析及估测  

Correlation Analysis and Estimation of Chlorophyll Content with Color Characteristics for Ginkgo biloba Under Water-Nitrogen Interactions

作　　者：裘赛铤 朱兴洲[1] 岳子龙 陶心语 周凯 QIU Sai-ting;ZHU Xing-zhou;YUE Zi-long;TAO Xin-yu;ZHOU Kai(Co-Innovation Center for Sustainable Forestry in Southern China,Nanjing Forestry University,Nanjing 210037,Jiangsu,China)

机构地区：[1]南京林业大学南方现代林业协同创新中心,江苏南京210037

出　　处：《西北林学院学报》2024年第4期15-23,共9页Journal of Northwest Forestry University

基　　金：国家自然科学基金项目(32101521);江苏省高等学校基础科学(自然科学)研究项目(21KJB220003)。

摘　　要：叶绿素含量是植物生长状况的重要指标。传统的测量叶绿素的方法费时费力,会对植物造成损伤。近年来,数字图像处理技术在估测植物叶绿素含量方向上取得了较好的进展,但针对银杏等经济林木的研究依旧偏少。以不同水氮互作条件下的2年生银杏幼苗为研究对象,使用数字扫描仪采集银杏幼苗叶片图像,利用数字图像技术构建颜色特征参数,结合相关性分析初筛出对叶绿素显著相关的颜色特征参数,并进一步基于高斯过程回归(gaussian process regression, GPR)和偏最小二乘回归(partial least squares regression, PLSR)优选建模中最为重要的颜色特征参数,建立基于银杏叶片颜色特征参数的叶绿素含量估测模型。结果表明,叶绿素含量随着施氮水平和水分处理水平的上升总体上呈现逐渐提高而后缓慢下降的趋势。基于单一颜色参数建立的单变量回归模型(R^(2)=0.01~0.72)预测精度总体上低于使用高斯过程回归(R^(2)=0.79~0.81)和偏最小二乘法(R^(2)=0.75~0.77)的模型。其中,GPR-BAT模型和PLSR-VIP模型都筛选出了对叶绿素敏感的R、G颜色特征参数;GPR模型的表现总体上优于PLSR模型,特别是在使用GPR-BAT优选颜色参数建模时表现最佳(R^(2)=0.81)。基于GPR-BAT优选颜色参数构建的GPR模型效果最佳,可准确估测银杏叶片叶绿素含量,为银杏生产的精确管理和监测银杏生长状况提供技术支撑。Chlorophyll content is an important indicator of plant growth status.Traditional methods for measuring chlorophyll are time-consuming and laborious,which can cause damage to plants.Recently,digital image processing technology has made great progress in estimating plant chlorophyll content,but there is still limited research focusing on non-commercial tree forest such as ginkgo(Ginkgo biloba).In this study,2-year-old ginkgo seedlings under different water-nitrogen interaction conditions were used as the research objects,and the leaf images of ginkgo seedlings were acquired by digital scanner.The color feature parameters were constructed by using digital image technology,and the color parameters that were significantly related to chlorophyll were screened out by correlation analysis,and the most important color parameters were selected based on Gaussian process regression(GPR)and partial least squares regression(PLSR)to establish a model for the estimation of chlorophyll content based on the color parameters of ginkgo leaves.The results showed that the chlorophyll content gradually increased with the increase of nitrogen application level and water treatment level,and then decreased slowly.The prediction accuracy of the univariate regression model based on a single color parameter(R^(2)=0.01-0.72)was generally lower than that of the model using GPR(R^(2)=0.79-0.81)and PLSR(R^(2)=0.75-0.77)methods.Among them,the GPR-BAT model and PLSR-VIP model both screened out the chlorophyll-sensitive R and G color characteristic parameters.GPR models performed better than PLSR models overall,especially when modeled using GPR-BAT preferred color parameters(R^(2)=0.81).The GPR model that is built on the basis of optimal color parameters(selected from the GPR-BAT method)performs best,which can accurately estimate the chlorophyll content of ginkgo leaves,and provide technical support for the accurate management of ginkgo production and monitoring the growth of ginkgo.

关 键 词：银杏 水氮互作 叶绿素 颜色特征 高斯过程回归 偏最小二乘回归 

分 类 号：S792.95[农业科学—林木遗传育种]