摘要
为解析宁夏滴灌玉米冠层图像参数与果穗形态参数间的内在联系,提出了一种采用作物冠层图像特征参数拟合玉米果穗生长发育动态的数学方法,建立玉米灌浆期果穗发育动态估算模型,实现了基于作物冠层数字图像处理技术的玉米果穗形态无损监测。用手机相机获取不同氮素处理下滴灌玉米灌浆期的冠层图像,提取玉米灌浆期冠层图像特征参数,测定玉米穗长、穗粗和穗体积等形态参数;运用R语言进行相关性分析,其中归一化冠层覆盖系数(Cc)与玉米果穗形态参数相关性高,运用Origin软件建立Cc与果穗形态参数间的估测模型,通过R2、RMSE和nRMSE评价估测模型的精度。结果表明,Cc与玉米穗长、穗粗、穗体积等形态参数均满足指数函数关系,其中Cc与穗长的预测精度最高,决定系数R2达到0.714,与穗粗的预测精度次之,R2为0.601,与穗体积的R2为0.575。由模型检验与评价结果可知,Cc与玉米果穗形态各参数间精度较高,其中R2均不小于0.523,穗体积RMSE的值均不大于68.986 cm^3,nRMSE均不大于33.621%。这表明基于冠层图像归一化覆盖系数的玉米果穗生长发育动态的估算具有一定的实用性,可为果穗形态参数估算和大面积玉米无损监测提供参考。
The main objective was to analyze the relationship between canopy image parameters and ear morphology parameters of drip irrigated maize in Ningxia.A mathematical method for fitting the growth characteristics of corn ear with crop canopy image parameters was proposed.The dynamic estimation model of ear development during corn filling stage was established to realize the non-destructive monitoring of corn ear morphology based on crop canopy digital image processing technology.The mobile phone camera technology was used to obtain the canopy image of the drip irrigated maize under different nitrogen treatments,and the canopy image parameters of maize filling stage were extracted to determine the morphological parameters such as ear length,ear diameter and ear volume.Correlation analysis was carried out by using R language to find out the normalized canopy cover factor(Cc)as the image feature parameter with high correlation between maize and ear morphological parameters.The Origin software was used to establish an estimation model between Cc and ear morphological parameters,and the accuracy of the model was estimated by R2,RMSE and nRMSE.The results showed that the morphological parameters of Cc and ear length,ear diameter and ear volume of maize satisfied the exponential function,and the prediction accuracy of Cc and ear length was the highest,the coefficient of determination R2 was 0.714,and the prediction accuracy of ear diameter was the second with R2 of 0.601 and R2 of ear volume was 0.575.From the model test and evaluation results,it can be seen that the accuracy of Cc and maize ear morphology parameters was higher,and the R2 was no less than 0.523,the ear volume RMSE value did not exceed 68.986 cm^3,and nRMSE was no more than 33.621%.Therefore,the estimation results of maize ear growth and development based on canopy image normalized coverage coefficient Cc had certain practicality,which can provide reference for estimation of ear shape parameters and non-destructive monitoring of large area maize.
作者
贾彪
贺正
王锐
孙权
王掌军
刘根红
JIA Biao;HE Zheng;WANG Rui;SUN Quan;WANG Zhangjun;LIU Genhong(School of Agriculture, Ningxia University, Yinchuan 750021, China)
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2020年第3期138-145,共8页
Transactions of the Chinese Society for Agricultural Machinery
基金
宁夏自然科学基金项目(2019AAC03068)
宁夏区重点研发计划项目(2019BBF03009,2018BBF02004)
国家自然科学基金项目(31560339)
西夏区科技局项目(2018XXKJ01)。
关键词
玉米
冠层图像
穗形态参数
氮素
估算模型
归一化冠层覆盖系数
maize
canopy images
ear phenotype parameters
nitrogen
estimation model
normalized canopy cover
