基于图像处理的茶树含氮量诊断方法初步研究  

A Preliminary Study on Nitrogen Diagnosis of Tea Plant Based on Image Processing

作　　者：刘连忠[1,2] 李孟杰 林源丰 张正竹 张嫚嫚 Liu Lianzhong;Li Mengjie;Lin Yuanfeng;Zhang Zhengzhu;Zhang Manman(School of Information and Computer,Anhui Agricultural University,Hefei 230036,China;State Key Laboratory of Tea Plant Biology and Resource Utilization,Hefei 230036,China;School of Tea and Food Science and Technology,Anhui Agricultural University,Hefei 230036,China)

机构地区：[1]安徽农业大学信息与计算机学院,安徽合肥230036 [2]茶树生物学与资源利用国家重点实验室,安徽合肥230036 [3]安徽农业大学茶与食品科技学院,安徽合肥230036

出　　处：《北华大学学报（自然科学版）》2019年第1期114-120,共7页Journal of Beihua University（Natural Science）

基　　金：茶树生物学与资源利用国家重点实验室开放基金课题(SKLTOF20160202);安徽省教育厅高校自然科学研究项目重点项目(KJ2017A151)

摘　　要：为快速获取茶树的氮营养素水平,基于图像处理技术研究茶树氮营养素诊断方法.采摘不同施氮地块同一生长部位的茶树叶片,在白色背景、自然光下用数码相机对叶片拍照,用化学法测得叶片含氮量.利用Open CV计算机视觉库编写图像处理程序,从原始图像中分割出叶片图像,提取叶片的R,G,B均值,并计算归一化的NRI,NGI,NBI及H,S,I均值,另选取R/G,B/G,B/R,R/(R+G-B),G/(R+G-B),B/(R+G-B)构成15个颜色特征参数,使用SPSS软件对颜色特征参数和含氮量进行回归分析,在误差分析基础上建立茶树氮营养素估算模型.将该模型运用于验证样本,得到含氮量估算值与实测值的平均相对误差为9. 28%,RMSE为0. 265.结果表明,该模型可用于茶树氮营养素水平的快速诊断.In order to obtain the nitrogen nutrition status of tea plant,a nitrogen diagnosis method was studied based on image processing technology.Leaves of tea plant with different nitrogen fertilizing in the field were picked as experimental material,and photographed on a white background and natural light with digital camera.Then the nitrogen content of leaves was measured by using Kjeldahl method.An image processing program of tea leaves was developed by using OpenCV,a computer vision library.By the program,leaf images were segmented,and fifteen color features of tea leaves were obtained,including R,G,B mean values,normalized NRI,NGI,NBI,H,S,I values,and R/G,B/G,B/R,R/(R+G-B),G/(R+G-B),B/(R+G-B).By SPSS,regression analysis of color features of tea leaves and nitrogen values were carried out to establish a nitrogen estimation model of tea plant based on error analysis.Applying the model to the test samples,the estimated values of nitrogen content were obtained with average relative error 9.28%,and RMSE 0.265.The result shows that the constructed nitrogen estimation model can be used for rapid nitrogen diagnosis of tea plant.

关 键 词：茶树 生长监测 图像处理 氮营养素诊断 精准施肥 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] S126[自动化与计算机技术—计算机科学与技术]