基于无人机可见光谱平台的烤烟氮素营养诊断

引用本文

孙志伟, 王晓琳, 张启明, 苑举民, 张爽, 闫慧峰, 王树声. 基于无人机可见光谱平台的烤烟氮素营养诊断[J]. 光谱学与光谱分析, 2021,41(2): 586-591.
SUN Zhi-wei, WANG Xiao-lin, ZHANG Qi-ming, YUAN Ju-min, ZHANG Shuang, YAN Hui-feng, WANG Shu-sheng. Diagnosis of Nitrogen Nutrition in Flue-Cured Tobacco Based on UAV Visible Spectrum Platform[J]. Spectroscopy and Spectral Analysis, 2021,41(2): 586-591.
Doi:10.3964/j.issn.1000-0593(2021)02-0586-06 复制到剪切板

Permissions

《光谱学与光谱分析》期刊社所有

基于无人机可见光谱平台的烤烟氮素营养诊断

孙志伟¹, 王晓琳¹, 张启明², 苑举民², 张爽¹, 闫慧峰^1,^*, 王树声^1,^*

1.中国农业科学院烟草研究所/农业农村部烟草生物学与加工重点实验室, 山东青岛 266101

2.江西省烟草科学研究所, 江西南昌 330025

*通讯作者 e-mail: yanhuifeng@caas.cn; wangshusheng@caas.cn

作者简介: 孙志伟, 1995年生, 中国农业科学院烟草研究所硕士研究生 e-mail: 732667191@qq.com

收稿日期: 2019-12-27 接受日期: 2020-03-16

基金: 国家重点研发计划项目(2018YFD0201100), 中央级公益性科研院所基本科研业务费专项(1610232019004), 中国烟草总公司江西省公司科技项目(2016.01.006)资助

摘要

利用不同氮肥用量田间试验, 分析基于无人机平台的可见光谱诊断技术对烟草氮素营养进行无损评估预测的可行性, 明确该技术的最佳颜色参数和方程模型。 2018年在江西省安福县开展田间试验, 设置不同氮肥用量, 分别为0, 45, 90, 135, 180和300 kg N·ha^-1, 于移栽后47 d(团棵期)、移栽后83 d(旺长后期)和移栽后116 d(下部叶成熟期), 利用无人机获取冠层RGB色彩数字图像, 同时采集植株样品分析地上部生物量、叶片生物量、地上部氮浓度、叶片氮浓度、叶片SPAD值等氮营养状况指标, 对冠层数字图像进行数字化分析, 获得颜色指标值, 通过颜色指标与烟草氮营养状况指标的相关性分析, 筛选适宜的颜色指标并建立氮营养诊断方程。利用不同地块的氮肥用量试验, 对氮营养诊断方程拟合精度进行验证。试验结果表明, 旺长后期各处理间冠层图像的颜色标准值存在显著差异, 团棵期与下部叶成熟期不存在显著差异。在10个颜色指标中, NRI, NGI, G/R, G/(R+B), (G-R)/(R+G+B)和ExG与5个烤烟氮素营养指标均达到极显著相关( p<0.01)。在归一化颜色指标体系、比颜色指标体系和归一化差分颜色指标体系中选择潜在的最佳颜色参数指标分别为NGI, G/R和ExG。根据不同类型的回归分析结果, 确定指数回归作为地上部生物量和叶片生物量的预测模型, 线性回归作为地上部氮浓度、叶片氮浓度及叶片SPAD值的预测模型。对潜在的最佳指标进行验证性筛选, G/R对地上部氮浓度和叶片氮浓度的RMSE值分别为0.375 1%和0.249 1%, 明显低于NGI和ExG, 预测精度最高。用G/R值表示的地上部生物量、叶片生物量、地上部氮浓度、叶片氮浓度、 SPAD值预测方程分别为 Y=21.785e^1.3502G/R, Y=4.057 9e^{1.937 3G/R}, Y=5.039 9G/R-3.333 2, Y=4.281 4G/R-3.802 9, Y=40.168G/R-28.188。因此, 基于无人机平台的可见光谱诊断技术在烤烟氮素营养诊断方面具有应用潜力, 评估最佳时期为旺长后期, 最佳预测参数为G/R值。

关键词: 烤烟; 冠层可见光谱; 氮营养诊断; 颜色指标; 方程模型; 生育期

中图分类号:O433.4 文献标志码:A

Diagnosis of Nitrogen Nutrition in Flue-Cured Tobacco Based on UAV Visible Spectrum Platform

SUN Zhi-wei¹, WANG Xiao-lin¹, ZHANG Qi-ming², YUAN Ju-min², ZHANG Shuang¹, YAN Hui-feng^1,^*, WANG Shu-sheng^1,^*

1. Tobacco Research Institute of Chinese Academy of Agricultural Sciences, Key Laboratory of Tobacco Biology and Processing, Ministry of Agriculture and Rural Affairs, Qingdao 266101, China

2. Jiangxi Institute of Tobacco Science, Nanchang 330025, China

*Corresponding authors

Abstract

The feasibility of non-destructive assessment and prediction of tobacco nitrogen nutrition based on visible spectrum diagnostic technology on UAV platform was analyzed by field experiments with different nitrogen dosage, and the optimum color parameters and model of the technology were defined. Field experiments were carried out in Anfu, Jiangxi Province in 2018. Different nitrogen dosages were set at 0, 45, 90, 135, 180 and 300 kg N·ha^-1. At 47 days after transplantation (cluster stage), 83 days after transplantation (flourishing late stage) and 116 days after transplantation (mature stage of lower leaves), digital images of canopy RGB color were obtained by UAV. At the same time, plant samples were collected to analyze aboveground biomass and leaf biomass. Digital analysis of canopy images was carried out to obtain the value of color indices. Through correlation analysis between color indices and tobacco nitrogen nutrition indices, appropriate color indices were screened and nitrogen nutrition diagnosis equation was established. The fitting accuracy of the diagnostic equation of nitrogen nutrition was validated by the experiment of nitrogen consumption in different plots. The results showed that the standard color values of canopy images were significantly different between different treatments in the later stage of flourishing growth, and there was no significant difference between the clustering stage and the maturity stage of lower leaves. Among the 15 color indexes, NRI, NGI, G/R, G/(R+B), (G-R)/(R+G+B) and ExG were significantly correlated with nitrogen nutrition indexes of 5 flue-cured tobacco varieties ( p<0.01). According to the screening method of canopy color index, NGI, G/R and ExG are the potential optimal color parameters in the normalized color index system, the ratio color index system and the normalized difference color index system. According to different types of regression analysis results, exponential regression was determined as the prediction model of aboveground biomass and leaf biomass, and linear regression as the prediction model of aboveground nitrogen concentration, leaf nitrogen concentration and leaf SPAD value. The RMSE values of G/R for aboveground nitrogen concentration and leaf nitrogen concentration were 0.375 1% and 0.249 1%, respectively, significantly lower than NGI and ExG, with the highest prediction accuracy. The prediction equations of above-ground biomass, leaf biomass, above-ground nitrogen concentration, leaf nitrogen concentration and SPAD value expressed by G/R value are Y=21.785e^{1.350 2G/R}, Y=4.057 9e^{1.937 3G/R}, Y=5.039 9G/R-3.333 2, Y=4.281 4G/R-3.802 9, Y=40.168G/R-28.188. Therefore, the UVAs platform-based visible spectrum diagnosis technology has application potential in the nitrogen nutrition diagnosis of flue-cured tobacco. The best evaluation period is the flourishing later stage, and the best prediction parameter is G/R value.

Keyword: Flue-cured tobacco; Canopy visible spectrum; Nitrogen diagnosis; Color index; Equation model; Growth period

文章图片

引言

氮素对烟草生长、品质及产量的影响极大。想要实时把握氮肥用量从而得到理想的烟叶产质, 田间氮素的快速诊断成为关键。传统的烤烟田间氮素诊断方法需要破坏性取样, 实时性差, 耗时长, 方法滞后。近年来, 非破坏性诊断工具在氮素快速诊断中应用越来越广泛。

无人机遥感具有方便灵活、使用成本低、响应快速、厘米级数据获取等优点^[1], 随着无人机(unmanned aerial vehicle, UAV)技术在农业领域的广泛应用, 使用无人机遥感(unmanned aerial vehicle remote sensing, UA CRS)对农作物进行大规模的氮素营养监测成为热点和趋势。毛智慧等证明玉米冠层植被指数与SPAD值有很强的相关性, 利用无人机多光谱技术可以进行高精度的叶绿素含量预测。大量研究显示, 农作物冠层氮质量分数与光谱反射率之间存在着显著的相关性^[2]。王远等^[3]利用图像分割技术提取的水稻红光标准化值NRI与SPAD值、叶片氮含量之间的相关性达到极显著水平。红光标准化值NRI可作为水稻和冬油菜氮素营养诊断的最佳冠层图像色彩参数指标^{[4, 5]}。李红军等^[6]发现返青期小麦反光叶面的G/R与NRI色彩参数与小麦的氮素营养状况相关性较好。刘昌华等^[7]研究也表明植被指数可以用来准确地估测冬小麦地上部生物量和植株吸氮量。

目前, 将无人机图像技术应用在烤烟氮素诊断的相关研究尚少, 本文旨在探讨通过无人机在田间能否进行氮素营养诊断, 确定烤烟快速诊断的最佳时期, 以及能否通过相关性分析筛选出具有代表性的图像色彩参数指标, 最终确定烤烟氮素营养快速诊断的方程模型, 为烤烟氮素营养无损诊断建立理论基础。

1 实验部分

1.1 试验设计

试验于2018年在江西省安福县进行。

进行氮营养诊断方程构建的试验地块位于E: 114.378, N: 27.380, 基础土壤地力为: 土壤pH值6.74, 有机质含量35.10 g· kg^-1, 碱解氮含量228.85 mg· kg^-1, 有效磷(P)含量137.10 mg· kg^-1, 速效钾(K)含量76.49 mg· kg^-1。烤烟品种为云烟87。设置6个不同氮肥用量的处理, 分别为N0: 0 kg N· ha^-1, N1: 45 kg N· ha^-1, N2: 90 kg N· ha^-1, N3: 135 kg N· ha^-1, N4: 180 kg N· ha^-1, N5: 300 kg N· ha^-1。每个处理的磷肥用量和钾肥用量均为135 kg P₂O₅· ha^-1和405 kg K₂O· ha^-1, 分别使用硝酸铵钙(15.5-0-0)、钙镁磷肥(0-12-0)和硫酸钾(0-0-50)做氮磷钾肥。试验设置3次重复, 每个小区4行× 15 m, 行距1.2 m, 株距0.45 m。肥料做基肥一次性施入。移栽时间为3月3日、移栽的株行距分别为1.2和0.45 m。其他田间管理措施均按照当地实际生产一致。

进行方程验证的试验地块位于E: 114.348, N: 27.337。烤烟品种为云烟87。设置5个不同氮肥用量的处理, 分别为0, 45, 90, 135和180 kg N· ha^-1。每个处理的磷肥用量和钾肥用量均为135 kg P₂O₅· ha^-1和405 kg K₂O· ha^-1。每个处理2行× 15 m, 行距1.2 m, 株距0.45 m。肥料做基肥一次性施入。肥料种类、移栽时间、移栽的株行距和其他田间管理措施均和进行氮营养诊断方程构建的试验地块相同。

1.2 测定项目

1.2.1 烤烟氮营养状况指标

取样的时期分别为移栽后47 d(团棵期)、移栽后83 d(旺长后期)和移栽后116 d(下部叶成熟期)。每次取样每个小区选取具有代表性的2株, 分为叶片和茎两部分, 分别在105 ℃下杀青30 min后烘干, 记录干重, 计算地上部生物量和叶片生物量。样品粉碎后采用H₂SO₄-H₂O₂消化, 利用全自动凯式定氮仪(Foss, Kjeltec8400)测定含氮量, 计算地上部氮浓度与叶片氮浓度。

每小区固定5株, 每次取样时测定所有叶片的SPAD值。平均后为叶片SPAD值。

1.2.2 冠层图像的获取

在晴天上午10点, 使用无人机(DJI, Phantom 3 Standard)在距地面30 m高度垂直拍摄整个试验地块的冠层图像。拍摄图像的分辨率为4 000× 3 000, 储存格式为JPEG。

1.2.3 数字图像的处理

获得的数字图像根据小区分布进行分割, 分割后的图像采用Adobe Photoshop CC(14.0× 32)软件, 使用直方图程序获取各小区的冠层图像的红光值R(redness intensity)、绿光值G(greenness intensity)和蓝光值B(blueness intensity)。选取3类颜色指标类型的15个颜色指标, 颜色指标分类及计算方法如表1所示。

表1 颜色指标分类及计算方法 Table 1 Classification and calculation of color indexes

1.3 数据分析方法

1.3.1 氮素营养诊断方程模型验证方法

通过方程计算验证地块的氮营养诊断指标预测值, 选择均方根误差(root mean square error, RMSE)评价预测值和实测值的拟合效果。均方根误差计算公式为

$RMSE = \sqrt[]{\frac{1}{n - 1} \overset{n}{\sum_{i = 1}} (y_{i} - y'_{i})^{2}}$

式中y'_i为样本i用标准方法测定的实际值。

1.3.2 数据处理

数据的处理采用Microsoft Excel 2016及SPSS 19。

2 结果与讨论

2.1 不同时期烤烟冠层图像色彩参数指标差异

不同时期不同处理的冠层图像的颜色标准值如表2所示。团棵期(移栽后47 d)和下部叶成熟期(移栽后116 d), 各处理冠层图像R, G和B值均没有显著差异, 这与视觉观察结果一致。旺长后期(移栽后83 d), 各处理间冠层图像的颜色标准值存在显著差异, 其中R值和G值随着氮肥施用量的增加逐步减小, B值在不同处理间差异不大; 在旺长后期时, 冠层图像在视觉观察上也存在叶色的明显差别。因此, 旺长后期(移栽后83 d)为利用无人机拍摄冠层图像进行烤烟氮氮素营养诊断的最佳时期。

表2 不同生育期各处理冠层图像颜色标准值的差异性分析 Table 2 Difference analysis of canopy image color standard values in different growth stages

2.2 烤烟冠层图像色彩参数指标的筛选

对旺长后期烤烟冠层3类10个色彩指标与烤烟氮素营养指标分别做相关性分析, 结果如表3所示。在平均颜色指标体系中, NRI和NGI分别与5个烤烟氮素营养指标极显著负相关和极显著正相关, NBI与地上部生物量和叶片生物量无显著相关性, 与地上部氮浓度、叶片氮浓度和叶片SPAD值显著正相关。在比颜色指标体系中, G/R和G/(R+B)与5个氮素营养指标均为极显著正相关; G/B与氮素营养指标相关性不高。在归一化差分颜色指标体系中, (G-R)/(R+G+B)和ExG与5个氮素营养指标均为极显著正相关; (R-B)/(R+G+B) 与4个烤烟氮素营养指标极显著相关, 而(G-B)/(R+G+B) 与氮素营养指标相关性不高。根据相关性, 在3类色彩指标体系中选择NGI, G/R和ExG作为潜在的最佳色彩参数指标。

表3 烤烟冠层图像色彩参数指标与烟株氮素营养指标的相关系数 Table 3 Interdependency between canopy color indexes and tobacco nitrogen indexes

2.3 烤烟氮素营养诊断方程构建

构建冠层色彩参数指标与氮营养指标的关系方程是烤烟氮素营养诊断的核心问题。采用线性和非线性模型(对数、二次、幂和指数), 对潜在的最佳色彩参数与氮素营养指标进行回归分析(图1)。结果表明, 五种回归模型下冠层图像色彩参数指标NGI, G/R和ExG与氮素营养指标间均表现出良好的相关性, 达到极显著水平。其中NGI和ExG与地上部生物量及叶片生物量的回归结果优于G/R的回归结果, 且幂函数回归和指数函数回归优于线性回归、二次回归和对数函数回归; G/R与地上部氮浓度、叶片氮浓度和叶片SPAD值的回归结果优于NGI和ExG的回归结果。根据相关系数的比较, 分别选择指数回归做为地上部生物量和叶片生物量的预测模型, 线性回归作为地上部氮浓度、叶片氮浓度及叶片SPAD值的预测模型。

	Figure Option View Download New Window
	图1 冠层色彩参数指标与氮营养指标的回归分析Fig.1 Regression analysis between canopy color indexes and tobacco nitrogen indexes

2.4 烤烟氮素营养诊断方程模型检验

选择相同生育期的不同地块进行氮营养诊断方程的验证分析, 不同图像色彩指标验证分析的结果如图2所示。验证结果表明, 基于颜色指标G/R建立的诊断方程模型, 地上部氮浓度及叶片氮浓度的真实检测值与方程预测值均较吻合, 重合度高。经计算NRI, G/R, ExG与地上部氮浓度的RMSE值分别为0.789 9%, 0.375 1%和0.788 1%, 与叶片氮浓度的RMSE分别为0.549 2%, 0.249 1%和0.549 0%。 RMSE反映了预测数据偏离真实值的程度, 值越小, 表示该模型预测精度越高。三个颜色指标中以G/R的RMSE值为最小, 预测精度最高, 应用G/R为参数的氮营养诊断方程的相比NRI与ExG更具有说服力。故选择G/R作为烤烟氮素营养指标预测的最佳图像色彩指标具有较高的代表性。

	Figure Option View Download New Window
	图2 以颜色指标为自变量的烤烟地上部氮浓度和叶片氮浓度预测值与实际值的关系Fig.2 Relationship between predicted and actual nitrogen concentrations in shoot and leaf of flue-cured tobacco using color index as independent variables

氮素作为影响烤烟生长发育和产量形成的必需元素之一, 在生产实际中, 氮肥施用过量仍然非常普遍。当地农民氮肥施用量一般在120~165 kg N· ha^-1之间^[13], 本研究设置了6个不同氮肥用量的处理(0, 45, 90, 135, 180和300 kg N· ha^-1), 通过分析不同氮素处理下各颜色指标与烤烟地上部生物量、叶片生物量、地上部氮浓度、叶片氮浓度、 SPAD值之间的关系, 筛选了NRI, G/R和ExR为无人机诊断烤烟氮素营养状况的备选指标, 继而通过对比不同方程模型间差异及RMSE值, 建立了基于G/R的烤烟旺长后期氮素营养指标线性方程模型。对“ 缺氮— 氮优化— 氮过剩” 之间应用无人机图像技术进行烤烟氮素营养诊断提供了一种方法和思路。本研究结果是在其他条件相同(磷钾肥用量充足且一致、拍照时间统一、拍摄高度角度保证一致等)的情况下得出的, 而烤烟不同生育期冠层颜色变化受很多因素影响, 如生长环境、养分类型及用量等, 因此, 探索不同条件下烤烟冠层颜色指标的差异机理及可行性成为今后工作的重点。同时, 本试验所有氮肥均做基肥一次性施用, 作物氮素营养无损诊断的核心是建立推荐追肥体系, 并以此进行施肥推荐。因此, 建立基于上述试验结果的不同生育期临界G/R值及推荐施氮量, 继而以追肥的方式提供氮肥, 仍需要更多、更广泛的试验来验证。相信随着精准农业和先进农业机械的快速发展, 该技术的应用将是作物无损营养诊断与推荐施肥的重要发展方向。

3 结论

通过无人机冠层图像分析监测烤烟氮素营养, 结果如下:

(1)对不同生育期冠层图像分析表明, 氮素营养诊断最佳的生育期为旺长后期。 (2)三类指标中筛选出NRI, G/R和ExG作为备选指标, 与营养指标相关性达到极显著水平, 证明在烤烟中通过无人机数字图像指标来预测营养指标的状况成为可能。 (3)五种回归模型下, 指数回归作为地上部生物量和叶片生物量的预测模型, 线性回归做为地上部氮浓度、叶片氮浓度及叶片SPAD值的预测模型。 (4)验证结果表明G/R对地上部氮浓度与叶片氮浓度的线性预测值与真实值更为接近, RMSE值更小, 我们最终选择G/R作为烤烟无人机氮素营养预测的最佳色彩指标。地上部生物量、叶片生物量、地上部氮浓度、叶片氮浓度、 SPAD值的预测方程为Y=21.785e^{1.350 2G/R}, Y=4.057 9e^{1.937 3G/R}, Y=5.039 9G/R-3.333 2, Y=4.281 4G/R-3.802 9, Y=40.168G/R-28.188。

参考文献

文献列表

[1]	WANG Xiao-qin, WANG Miao-miao, WANG Shao-qiang, et al(王小钦, 王苗苗, 王绍强, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2015, 31(5): 152. [本文引用:1]
[2]	Yao Y, Miao Y, Huang S, et al. Agronomy for Sustainable Development, 2012, 32(4): 925. [本文引用:1]
[3]	WANG Yuan, WANG De-jian, ZHANG Gang, et al(王远, 王德建, 张刚, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2012, 28(17): 131. [本文引用:1]
[4]	LI Lan-tao, ZHANG Meng, REN Tao, et al(李岚涛, 张萌, 任涛, 等). Journal of Plant Nutrition and Fertilizer(植物营养与肥料学报), 2015, 21(1): 259. [本文引用:1]
[5]	WEI Quan-quan, LI Lan-tao, REN Tao, et al(魏全全, 李岚涛, 任涛, 等). Scientia Agricultura Sinica(中国农业科学), 2015, 48(19): 3877. [本文引用:1]
[6]	LI Hong-Jun, ZHANG Li-zhou, CHEN Xi-ming, et al(李红军, 张立周, 陈曦鸣, 等). Chinese Journal of Eco-Agriculture(中国生态农业学报), 2011, 19(1): 155. [本文引用:1]
[7]	LIU Chang-hua, MA Wen-yu, CHEN Zhi-chao(刘昌华, 马文玉, 陈志超). Journal of Henan University of Technology·Natural Science Edition(河南理工大学学报·自然科学版), 2018, 37(3): 45. [本文引用:1]
[8]	CHEN Jia-yue, YAO Xia, HUANG Fen, et al(陈佳悦, 姚霞, 黄芬, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2016, 32(4): 163. [本文引用:1]
[9]	ZHANG Ling, CHEN Xin-ping, JIA Liang-liang, et al(张玲, 陈新平, 贾良良, 等). Journal of Plant Nutrition and Fertilizer(植物营养与肥料学报), 2018, 24(1): 261. [本文引用:1]
[10]	LI Jin-wei, GUAN He-qing, LIAO Gui-ping(李锦卫, 管鹤卿, 廖桂平). Agriculture Network Information(农业网络信息), 2010, 12: 15. [本文引用:1]
[11]	Yao X, Si H, Chen T, et al. Frontiers in Plant Science, 2018, 9: 1360. [本文引用:1]
[12]	LIU Hong-jian, ZENG Ai-ping, ZHENG Li-min(刘洪见, 曾爱平, 郑丽敏). Agricultural Network Information(农业网络信息), 2007, (12): 31. [本文引用:1]
[13]	TANG Xian-gan, LI Zu-zhang, HU Qi-feng, et al(唐先干, 李祖章, 胡启锋, 等). Chinese Tobacco Science(中国烟草科学), 2012, 33(3): 47. [本文引用:1]

2015

0.0

... 无人机遥感具有方便灵活、使用成本低、响应快速、厘米级数据获取等优点^[1], 随着无人机(unmanned aerial vehicle, UAV)技术在农业领域的广泛应用, 使用无人机遥感(unmanned aerial vehicle remote sensing, UA CRS)对农作物进行大规模的氮素营养监测成为热点和趋势 ...

2012

0.0

... 大量研究显示, 农作物冠层氮质量分数与光谱反射率之间存在着显著的相关性^[2] ...

2012

0.0

WANG

Yuan

, WANG

De-jian

, ZHANG

Gang

, et al(王远, 王德建, 张刚, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2012, 28(17): 131.

Segmenting canopy region effectively from the image captured by a digital camera is necessary for diagnosis of crop nutrition. An image segmentation thresholding method was proposed based on the difference of green and red channel (GMR) according to the difference of reflectance spectrum between vegetation and soil in visible band. The range of optimum threshold was from 10 to 20 during the rice growing stage, and the best threshold for shooting and booting stage was 10 and 20, respectively. The feature parameters of the segmentation of canopy images using this method are in good agreement with the SPAD readings, leaf nitrogen content, etc, in which the correlation coefficient between normalized redness intensity (NRI) and the two reached -0.87, -0.65, respectively. It demonstrates that this method is not only simple and accurate, but also applicable universally to other green crops.

利用数码相机对作物进行快速准确的营养诊断，需要对图像中作物冠层部分与非冠层部分进行有效的分割。该文依据绿色植被和土壤在可见光区域反射光谱的差异，提出了根据数字图像绿色通道和红色通道差值的大小设定阈值对图像进行分割的方法。阈值设定为10～20之间时对水稻冠层图像有较好的分割效果，拔节期和孕穗期获得最佳图像分割效果的阈值分别为10与20。分割后图像中提取的特征参数与SPAD值、叶片含氮量等指标间具有良好的相关关系，其中红光标准化值NRI与两者间的相关系数达到-0.87和-0.65。该方法能准确地分割水稻冠层图像，且简便易行，对绿色植被的图像分割具有普适性，有较高应用价值。

... 王远等^[3]利用图像分割技术提取的水稻红光标准化值NRI与SPAD值、叶片氮含量之间的相关性达到极显著水平 ...

2015

0.0

Lan-tao

, ZHANG

Meng

, REN

Tao

, et al(李岚涛, 张萌, 任涛, 等). Journal of Plant Nutrition and Fertilizer(植物营养与肥料学报), 2015, 21(1): 259.

【Objectives】 The spatial and temporal distribution of color indexes of canopy (G, NRI, NGI, NBI, G/R and G/B) and the indexes of N nutrition in rice plants were studied to determine the best color parameters and regression equations for nitrogen with a digital camera, and provide a theoretical basis and technical approach for monitoring plant nitrogen status of rice and precision management of nitrogen fertilization. 【Methods】 Field experiments were carried out from May to September 2013 in rice growing season at the Experimental Farm of Huazhong Agricultural University (30°28′ 08′′N, 114°21′36′′E). A two-line indica hybrid (Liangyou6326) was chosen as test cultivar. Four N treatments: N 0, 75, 150 and 225 kg/ha were designed and recorded as N0, N75, N150, N225, respectively. A color digital camera (D700, Nikon, Japan) with a resolution of 12.0 mega pixels was employed to capture color images of rice canopy at the tillering, jointing, booting and filling stage. The image resolution was 1936×1296 pixels of 14 bit for red, green and blue. The images of size 2.72 MB were transferred in joint photographic experts group (JPEG) format to a computer and processed with Adobe Photoshop 7.0 to extract color information for studying possibility of using digital image analysis method for plant N status diagnosis and for determining the best color parameters and regression equations in rice. 【Results】 Compared with N0, the biomass, leaf N content, plant total N content, N accumulation and canopy NDVI values of rice with N treatments at the tillering stage, jointing stage, booting stage and filling stage are averagely increased by 40.7%-98.0%, 42.4%-72.4%, 36.2%-85.3%, 125.5%-209.1% and 51.3%-60.6%, respectively, at mature stage, the average increase of yield are from 60.1% to 117.0%. Compared with other plant canopy color parameters, the normalized redness intensity (NRI), calculated as R/(R+G+B), is much better as it has better correlations with the index of N-nutrient in rice plants, rice yield and canopy NDVI values at the different growth stages. An integrated linear regression equation could be used for describing the relationship between NRI and leaf N content, plant total N content, N accumulation, canopy NDVI values, yield at different growing stages and nitrogen levels of rice. 【Conclusions】 The digital image processing technique is usable in detecting the N nutrition in rice and has advantages of fast, stable results, easy to apply, and non-destructive. Simultaneously, the color parameter NRI has a preferable interrelation with the index of N-nutrition and yield than others, thus NRI is suitable for rapid diagnosis of nitrogen nutrition, and the digital image processing technique method shows the potential of being used in fast rice nitrogen diagnosis without damage.

College of Resources and Environment,Huazhong Agricultural University/Key Laboratory of Arable Land Conservation (Middle and Lower Reaches of Yangtse River), Ministry of Agriculture, Wuhan 430070, China

【目的】研究田间试验条件下水稻不同生育期冠层图像色彩参数(G、 NRI、 NGI、 NBI、 G/R和G/B)及植株氮素营养指标(叶片含氮量、植株全氮含量、生物量、氮素累积量和冠层NDVI值)的时空变化特征,并分析两者间的相关性,确立水稻氮素营养诊断的最佳色彩参数和方程模型,为探明数码相机在水稻上的适宜性及精确诊断水稻氮素营养状况提供理论基础。【方法】于2013年5月~9月在湖北省武汉市华中农业大学试验基地(30°28′ 08′′N,114°21′36′′E)采用不同施氮处理的田间试验,以籼型两系杂交稻“两优6326”为供试作物,设置4个施氮水平: 0、 75、 150和225 kg/hm 2 (分别以N0、 N75、 150和N225表示),3次重复,随机区组排列。分别在水稻分蘖期、拔节期、孕穗期和灌浆期采用数码相机(Nikon-D700,1200万像素)获取水稻冠层图像,应用Adobe photoshop7.0软件直方图程序提取图像的红光值R、绿光值G和蓝光值B,研究数码相机进行水稻氮素营养诊断色彩参数,确定植株氮素营养指标诊断模型。【结果】较对照(N0)相比,分蘖期、拔节期、孕穗期和灌浆期3个施氮处理水稻地上部生物量、叶片含氮量、植株全氮含量、氮素累积量、冠层NDVI值和成熟期产量增幅分别平均为40.7%~98.0%、 42.4%~72.4%、 36.2%~85.3%、 125.5%~209.1%、 51.3%~60.6%和60.1%~117.0%,差异显著。水稻不同生育期各冠层数字化指标G、 NRI、 NGI、 NBI 、 G/R和G/B与上述氮素营养参数相关性差异较大,且以数字图像红光标准化值NRI表现最佳,建议作为应用数码相机进行水稻氮素营养诊断的最佳冠层图像色彩参数指标。进一步分析表明,可以用统一的线性回归方程来描述不同生育期、不同氮素水平下水稻植株氮素营养指标随冠层色彩参数NRI的变化模式。【结论】数码相机进行水稻氮素营养诊断测试结果稳定,具有快速、便捷、非破坏性等优点,冠层色彩参数NRI与水稻氮素营养指标和产量之间均表现出较好的相关性,满足氮素营养无损诊断的需求,对实时、快速监测水稻长势状况及氮素营养丰缺水平具有较高的可行性,有望发展成为新时期作物氮素营养无损诊断技术的潜力。

... 红光标准化值NRI可作为水稻和冬油菜氮素营养诊断的最佳冠层图像色彩参数指标^[4,5] ...

2015

0.0

WEI

Quan-quan

, LI

Lan-tao

, REN

Tao

, et al(魏全全, 李岚涛, 任涛, 等). Scientia Agricultura Sinica(中国农业科学), 2015, 48(19): 3877.

【Objective】 In order to provide a scientific basis for digital image processing technique in nitrogen nondestructive diagnosis of winter rapeseed, a field experiment was carried out to explore the feasibility of digital image processing technique in nitrogen nondestructive diagnosis, determine the best digital parameter and regression equation. 【Method】 A field experiment was conducted with different nitrogen application rates (0, 90, 180, 270 and 360 kg·hm -2 ). The pictures of winter rapeseed canopy were obtained with a digital camera, meanwhile, the conventional nitrogen diagnosis parameters of aboveground biomass, nitrogen concentration in leaf and leaf chlorophyll content at six-leaf period, ten-leaf period, bud period and blooming period, were determined and the correlations were analyzed. 【Result】 The red color intensity (R), normalized redness intensity (NRI) and the ratio of greenness and blueness intensity (G/B) had significant inverse correlations with the conventional N diagnosis parameters. But green color intensity (G), blue color intensity (B), the ratio of greenness and redness intensity (G/R), the ratio of blueness and redness intensity (B/R), the normalized greenness intensity (NGI) and the normalized blueness intensity (NBI) showed a significant positive correlation. Compared with other digital index, the normalized redness intensity (NRI) showed a prominently and very prominently inverse relationship with conventional N parameters and the regression equation with N application rate, aboveground biomass, leaf N concentration, chlorophyll in leaf, N uptake and N nutrition index is y (t·hm -2 )=-8.003 x +2.706, y (t·hm -2 )=-106.072 x +38.200, y (g·kg -1 )=-692.99 x +261.84, y (mg·g -1 )=-12.750 x +5.665, y (kg·hm -2 )=-4087.416 x +1414.274 and y =-27.198 x +9.812, the correlation values ( R 2 ) is 0.917 ** , 0.746 ** and 0.953 ** between measured and estimated leaf N concentration, chlorophyll in leaf and NNI , RMSE and RE were 0.821, 26.32%, 0.330, 28.57% and 0.228, 28.39%, respectively. 【Conclusion】Digital image processing technique can be used in nitrogen nondestructive diagnosis of winter rapeseed, the stage before bud period is crucial period with NRI as the index for the N nitrogen diagnosis using digital image processing technique, and NRI is the best diagnosis index with linear regression equation.

1 College of Resources and Environment, Huazhong Agricultural University, Wuhan 430070 2 Key Laboratory of Arable Land Conservation (Middle and Lower Reaches of Yangtse River), Ministry of Agriculture, Wuhan 430070 3 Wuxue Bureau of Agriculture, Wuxue 435400, Hubei

【目的】利用田间氮肥梯度试验探讨数字图像技术对冬油菜氮素营养无损评估预测的可行性,明确该技术的最佳数码参数和方程模型,为数字图像技术进行冬油菜氮素无损诊断提供依据。【方法】2013—2014年在湖北省武穴市开展不同施氮处理田间试验,以冬油菜为试验材料,设置不同氮素水平（0、90、180、270和360 kg·hm -2 ）,分别于六叶期、十叶期、蕾薹期和开花期,利用数码相机获取冠层数字图像数据,同时采集植株样品分析其生长特征值,研究其相关性并建立氮素营养参数的方程模型。利用 2014—2015年独立氮肥水平试验,对上述方程模型拟合精度进行验证并绘制1﹕1线性关系图。【结果】数字图像红光值（R）、红光标准化值（NRI）和绿光与蓝光比值（G/B）与冬油菜氮营养状况常规诊断指标地上部生物量、叶片氮浓度和叶绿素浓度等呈负相关关系,而绿光值（G）、蓝光值（B）、绿光与红光比值（G/R）、蓝光与红光比值（B/R）、绿光标准化值（NGI）和蓝光标准化值（NBI）则与上述指标呈正相关关系,红光标准化值（NRI）与其他数码参数相比能更好地表征冬油菜的氮素营养状况,蕾薹期红光标准化值NRI与氮肥用量、地上部生物量、叶片氮浓度、叶绿素浓度、氮素吸收量和氮营养指数之间的关系可分别用线性方程 y (t·hm -2 )=-8.003 x +2.706、 y (t·hm -2 )=-106.072 x +38.200、 y (g·kg -1 )=-692.99 x + 261.84、 y (mg·g -1 )=-12.750 x +5.665、 y (kg·hm -2 )=-4087.416 x +1414.274和 y =-27.198 x +9.812来表达,其相关性达到极显著水平。2014—2015年独立试验模型检验结果表明,叶片氮浓度、叶绿素浓度和氮营养指数实测值与预测值的决定系数 R 2 分别为0.917 ** 、0.746 ** 和0.953 ** ；均方根误差 RMSE 分别为0.821、0.330和0.228；相对误差 RE %分别为26.32%、28.57%和28.39%,模型预测精度较好。【结论】数字图像技术可以用于冬油菜氮素营养的评估预测,评估时期为蕾薹期（包括）之前均可,最佳预测参数为红光标准化值NRI,参数的最佳方程模型为直线方程函数。

... 红光标准化值NRI可作为水稻和冬油菜氮素营养诊断的最佳冠层图像色彩参数指标^[4,5] ...

2011

0.0

Hong-Jun

, ZHANG

Li-zhou

, CHEN

Xi-ming

, et al(李红军, 张立周, 陈曦鸣, 等). Chinese Journal of Eco-Agriculture(中国生态农业学报), 2011, 19(1): 155.

Easy, fast and cheap diagnosis of crop nitrogen status is a key to optimize nitrogen fertilization. Crop nitrogen status can be measured by relating it to color parameters retrieved form digital images. In this paper, color parameters of digital image of winter wheat under different levels of nitrogen supply were analyzed by using Adobe Photoshop and PCI Geomatics remote sensing image processing software and the results compared. Based on color differences, winter wheat image was divided into three parts (soil, leaves with and without reflection). Correlations among eight color parameters retrieved by PCI and Photoshop, SPAD values and total nitrogen content of wheat were analyzed. The results showed that G/R and R/(R+G+B) ratios of leaves with reflection were significantly correlated with wheat SPAD at reviving stage. R/(R+G+B) ratio of leaves with and without reflection was significantly correlated with wheat total nitrogen content at jointing stage. It was noted that improvements in image processing for getting color parameters of software were required. Color parameters obtained via image classification could increase the diagnosis accuracy of wheat nitrogen status.

简便、快速、经济地诊断作物氮素营养状况是实施氮肥用量调控的关键。利用数码相机对作物冠层进行拍照, 通过图像处理软件获得作物色彩参数, 根据色彩参数与作物氮素营养状况的关系可以对其氮素丰缺进行诊断。针对作物数字图像色彩参数的获取方法, 结合小麦多水平氮肥试验, 采用遥感软件PCI Geomatics的非监督分类功能, 将小麦图像分为土壤、反光叶面和不反光叶面, 与Adobe Photoshop 软件普通图像处理方法对照, 比较分析了小麦图像不同类别叶片的8 种色彩参数与SPAD 值及植株全氮的相关性。结果表明, 返青期小麦反光叶面的G/R 与R/(R+G+B)色彩参数能较好地反映小麦的氮素营养状况; 拔节期不反光叶面和反光叶面的R/(R+G+B)色彩参数与植株全氮相关性较好。利用普通图像处理软件获得色彩参数的方法有待改进, 图像分类后能够提高其色彩参数对作物氮素营养诊断的准确性。

... 李红军等^[6]发现返青期小麦反光叶面的G/R与NRI色彩参数与小麦的氮素营养状况相关性较好 ...

2018

0.0

... 刘昌华等^[7]研究也表明植被指数可以用来准确地估测冬小麦地上部生物量和植株吸氮量 ...

2016

0.0

CHEN

Jia-yue

, YAO

Xia

, HUANG

Fen

, et al(陈佳悦, 姚霞, 黄芬, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2016, 32(4): 163.

为探索基于数字图像处理技术的冬小麦氮素无损诊断图像评价指标及构建方法，设计拍摄2012-2014年度不同种植方案下冬小麦冠层图像，基于归一化的H分量K均值聚类分割算法提取基础颜色特征值，与同期叶片氮含量（leaf nitrogen content，LNC）进行线性拟合，调优并确定三原色分量最佳拟合系数，提出RGB空间下的颜色组合标准化指数（normalized color mix index，NCMI）。对比深绿色指数（dark green color index，DGCI）、红光标准化值（normalized redness intensity，NRI）和绿光与红光比值G/R发现，3个采样期NCMI与LNC的决定系数R2均高于3个对比指标，分别为0.77、0.79、0.94，均方根误差（root mean square error，RMSE）相较同期最低的指标，分别降低了0.18%、0.37%和1.67%；生选6号和扬麦18号NCMI与LNC的相关性，在一定冠层覆盖度下均优于其他3个指标；D2密度（3×106株/hm2）N1（纯氮150 kg/hm2）处理下NCMI效果明显优于其他3个指标，R2和RMSE较NRI分别改善了7.69%和4.11%，该研究可为一定冠层覆盖度下的冬小麦氮素营养诊断图像评价指标提供参考。

2018

0.0

2010

0.0

Jin-wei

, GUAN

He-qing

, LIAO

Gui-ping

(李锦卫, 管鹤卿, 廖桂平). Agriculture Network Information(农业网络信息), 2010, 12: 15.

介绍了利用计算机视觉技术测量油菜叶面积的原理和方法,并以纸重法为标准对比分析了传统直尺法与此方法在油菜三种类型叶片叶面积测量中的优劣性.结果表明,计算机视觉法在精度上明显高于直尺法,尤其在无柄叶测量中此方法的测量平均相对误差为3.09%,低于直尺法的10.49%,进一步根据标准测量值对直尺法测量无柄叶时的系数进行校正,发现系数选择0.75更为合适.

2018

0.0

2007

0.0

LIU

Hong-jian

, ZENG

Ai-ping

, ZHENG

Li-min

(刘洪见, 曾爱平, 郑丽敏). Agricultural Network Information(农业网络信息), 2007, (12): 31.

本文在图像处理技术的基础上,通过分析颜色特征与玉米氮营养之间的关系,选取了与玉米营养状况相关性高的颜色特征,建立了基于颜色特征的玉米氮营养诊断模型,并认为RGB颜色空间比其他的颜色空间更适合于描述和表达玉米营养状况.

2012

0.0

... ha^-1之间^[13], 本研究设置了6个不同氮肥用量的处理(0, 45, 90, 135, 180和300 kg N#cod#x000b7 ...