首页 | 简介 | 作者 | 编者 | 读者 | Ei(光盘版)收录本刊数据 | 网络预印版 | 点击排行前100篇 | English
张珏,田海清,李哲,李斐,史树德.基于数码相机图像的甜菜冠层氮素营养监测[J].农业工程学报,2018,34(1):157-163.DOI:10.11975/j.issn.1002-6819.2018.01.021
基于数码相机图像的甜菜冠层氮素营养监测
投稿时间:2017-06-22  修订日期:2017-12-10
中文关键词:  图像处理    模型  甜菜  数码相机  颜色特征参数  氮素营养监测
基金项目:国家自然科学基金项目(41261084);内蒙古自然科学基金项目(2016MS0346);国家现代农业产业技术体系专项基金项目(CARS-210402)
作者单位
张珏 1.内蒙古农业大学机电工程学院呼和浩特 010018
2. 内蒙古师范大学物理与电子信息学院呼和浩特 010020
 
田海清 1.内蒙古农业大学机电工程学院呼和浩特 010018
 
李哲 1.内蒙古农业大学机电工程学院呼和浩特 010018
 
李斐 3. 内蒙古农业大学草原与资源环境学院呼和浩特 010019
 
史树德 4. 内蒙古农业大学农学院呼和浩特 010019
 
摘要点击次数: 396
全文下载次数: 234
中文摘要:为探究数码相机监测甜菜冠层叶片氮素的可行性,2014年于内蒙古赤峰市松山区太平地镇采用不同种植方案设计了田间试验。利用数码相机获取甜菜冠层数字图像,基于灰度值的阈值分割方法提取冠层图像的红光值(R)、绿光值(G)和蓝光值(B),交互调优R、G、B单色分量权重,提出三原色权值调优方法,并挖掘出适宜于表征甜菜冠层LNC(leaf nitrogen content)的基础调优参数BOP(basic optimal parameter)和归一化调优参数NOP(normalized optimal parameter)。结果表明:采用常规方法选取的敏感颜色参数G/R、NRI(R/(R+G+B))与冠层LNC的相关系数分别为0.80和0.79,三原色权值调优方法确定的调优参数BOP、NOP与冠层LNC的相关系数分别为0.83和0.84,算法优化后提高了颜色参数与冠层LNC的相关性。对比常规参数和调优参数对冠层LNC的预测精度,结果显示调优参数BOP、NOP建立模型的预测精度均高于常规参数G/R、NRI,BOP预测模型的决定系数R2和均方根误差RMSE(root mean square error)分别为0.69和2.65,NOP预测模型的R2和RMSE分别为0.68和2.73。该研究表明,在大田自然光照条件下,借助数码相机实时、准确监测甜菜氮素营养丰缺水平具有较高的可行性,数字图像处理技术在作物营养无损诊断中存在很大的应用潜力。
Zhang Jue,Tian Haiqing,Li Zhe,Li Fei,Shi Shude.Nitrogen nutrition monitoring of beet canopy based on digital camera image[J].Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE),2018,34(1):157-163.DOI:10.11975/j.issn.1002-6819.2018.01.021
Nitrogen nutrition monitoring of beet canopy based on digital camera image
Author NameAffiliation
Zhang Jue 1. College of Mechanical and Electrical Engineering, Inner Mongolia Agricultural University, Hohhot 010018, China
2. College of Physics and Electronic Information, Inner Mongolia Normal University, Hohhot 010020, China
 
Tian Haiqing 1. College of Mechanical and Electrical Engineering, Inner Mongolia Agricultural University, Hohhot 010018, China
 
Li Zhe 1. College of Mechanical and Electrical Engineering, Inner Mongolia Agricultural University, Hohhot 010018, China
 
Li Fei 3. College of Grassland, Resources and Environment Science, Inner Mongolia Agricultural University, Hohhot 010019, China
 
Shi Shude 4. College of Agronomy, Inner Mongolia Agricultural University, Hohhot 010019, China
 
Key words:image processing  nitrogen  model  beet  digital camera  color characteristic parameter  nitrogen nutrition monitoring
Abstract: To explore the feasibility of monitoring nitrogen elements in beet canopy leaves by digital camera, field experiment with different planting schemes was carried out in Chifeng City, Inner Mongolia in 2014. Canopy digital images of beet grown under different nitrogen application rates were captured several times during the whole growth stage. The change trend of canopy LNC (leaf nitrogen content) under different nitrogen levels was analyzed. It was found that canopy LNC is relatively high in the middle period of rapid growth and later stage of sugar growth, and the canopy grow trend is from high to low during the whole stage. Canon EOS7D digital camera with the resolution of 5184×3456 was used for image acquisition. In order to keep the light source consistent and improve the comparability of the images captured in different stages, collection time was set at noon from 12:00 to 14:00, when the weather was clear and calm. The camera was 1.50 m above the beet canopy and had an included angle of 60° with the ground. Complete beet canopies of adjacent 2×2 plants were selected and the image was stored in JPEG format. In view of the difference of gray value between leaves and background, threshold segmentation method based on gray value was used to segment the soil, shading leaves, and numbered signs. After that, the color image with green leaves only was obtained, and the R (redness intensity), G (greenness intensity), and B (blueness intensity) values were extracted. Ten image feature parameters were chosen to analyze their correlation with monitoring evaluation index of canopy nutrition under different schemes, including 3 single color characteristic values (R, G and B), 4 linear combination parameters (G/R, G+B, R/B and R-B), and 3 linear combination parameters by standardized processing (R/R+G+B, G/R+G+B and B/R+G+B). It was found that different characterization ability exists among 3 single color parameters, and the correlation between the composite characteristic parameters and the canopy LNC has significant improvement compared with that between the single color parameter and the canopy LNC. Interactive tuning R, G and B tricolor component coefficients, and the method of primary color weight optimization was proposed. The BOP (basic optimal parameter) and NOP (normalized optimal parameter) were extracted to characterize nitrogen elements in beet canopy leaves. The results show that the 2 tuning parameters have great effect on correlation and fitting accuracy compared with color characteristic parameters G/R and NRI (normalized redness intensity) obtained by conventional methods. The correlation coefficients between G/R, NRI and canopy LNC were 0.80 and 0.79, respectively, and those between BOP, NOP and canopy LNC were 0.84 and 0.83, respectively. Comparing prediction accuracy, the R2 values of the model based on the tuning parameters BOP and NOP were 0.69 and 0.68, respectively, which were higher than the contrast indices G/R and NRI. The RMSE (root mean square error) values of the model based on the tuning parameters were 2.65 and 2.73, respectively, which were lower than the contrast indices. It can be seen that the method of primary color weight optimization is more efficient than the conventional color characteristic parameter selection method. The sensitive parameters affect the accuracy of crop nutrition monitoring, however, most studies choose sensitive parameters from the common, and researches are few in the constructing parameters method. This paper analyzed R, G and B primary weights, and proposed a general method for constructing color sensitive parameters. The optimization parameters BOP, NOP and coefficient optimization model were constructed, and the parameters weight range was analyzed and regulated, and then BOP and NOP were optimized. This study can provide a basis for the nutritional diagnosis of other crops, and also show it is possible to estimate the nutrition deficiency using digital camera under the conditions of field natural light. This indicates that conventional low-cost digital cameras can be used for determining nitrogen content in beet canopy leaves, and also offers a potentially inexpensive, fast, accurate and suitable tool for small farms.
查看全文   下载PDF阅读器

京ICP备06025802号-3

主办单位:中国农业工程学会 单位地址:北京朝阳区麦子店街41号

服务热线:010-59197076、59197077 传真: 邮编:100125 Email:tcsae@tcsae.org
本系统由北京勤云科技发展有限公司设计

京公网安备 11010502031390号