Study on estimation model of total nitrogen content in rice leaves based on image characteristics

doi:10.3969/j.issn.1004-1524.2020.12.15

Acta Agriculturae Zhejiangensis ›› 2020, Vol. 32 ›› Issue (12): 2232-2243.DOI: 10.3969/j.issn.1004-1524.2020.12.15

• Biosystems Engineering • Previous Articles Next Articles

Study on estimation model of total nitrogen content in rice leaves based on image characteristics

YANG Hongyun¹(), LUO Jianjun², SUN Aizhen², WAN Ying², YI Wenlong¹

1. School of Software, Jiangxi Agricultural University/Key Laboratory of Agricultural Information Technology, Jiangxi Higher Education, Nanchang 330045, China
2. School of Computer and Information Engineering, Jiangxi Agricultural University, Nanchang 330045,China

Received:2020-04-17 Online:2020-12-25 Published:2020-12-25

Abstract

Abstract:

In order to explore the quantitative description of the relationship between the external color, geometric shape characteristics of rice leaf growth and its total nitrogen content, it can quickly and accurately diagnose the nitrogen nutrition status of rice. In this study, the sensitive leaf positions for total nitrogen content estimation were screened out, and the models based on multiple linear regression and machine learning methods for total nitrogen content estimation of rice sensitive leaf positions were compared, which provided ideas and methods for establishing high-performance quantitative diagnosis model of nitrogen nutrition. The field experiment of rice was carried out in Chengxin farm, Nanchang City, Jiangxi Province from 2017 to 2018. The tested rice variety was LYP9. Four nitrogen application levels (0, 210, 300 and 390 kg·hm^-2 from low to high) were set. At the panicle differentiation stage and full heading stage of rice, 4 800 images of the first fully expanded leaf (1st leaf), the second fully expanded leaf (2nd leaf) and the third fully expanded leaf (3rd leaf) at the top of rice were obtained by scanning. Twenty-five scanning leaf color and geometric shape characteristics of rice were acquired by image processing technology, and the sensitive leaf positions in two periods were screened by multiple linear regression, and the total nitrogen content estimation model of rice sensitive leaf positions was established by machine learning method. The average relative errors of the length and width of rice leaves obtained by image processing method were 0.328% and 3.404% respectively, compared with the manual measurement; the 3rd leaf at the panicle differentiation stage and the 2nd leaf at the full heading stage were more sensitive than other leaf positions at the same period, and the 3rd leaf at the panicle differentiation stage was the most sensitive; the model of estimating the total nitrogen content in sensitive leaves of rice based on machine learning was slightly better than the multiple linear regression, and BP neural network was the best, RMSE_v=0.089, MRE_v=0.034,$R^{2}_{v}$=0.887 in the 3rd leaf at the panicle differentiation stage; RMSE_v=0.132, MRE_v=0.046,$R^{2}_{v}$=0.820 in the top 2nd leaf at the full heading stage. The 3rd leaf of the panicle differentiation stage and the 2nd leaf of the full heading stage were the most representative and more feasible to estimate the total nitrogen content, which can be used as an effective leaf position for nitrogen nutrition diagnosis of rice.

Key words: rice, leaf total nitrogen content, multiple linear regression, machine learning, support vector machine, BP neural network

CLC Number:

S126

YANG Hongyun, LUO Jianjun, SUN Aizhen, WAN Ying, YI Wenlong. Study on estimation model of total nitrogen content in rice leaves based on image characteristics[J]. Acta Agriculturae Zhejiangensis, 2020, 32(12): 2232-2243.

Figures/Tables 12

Fig.1 Comparison image of six channel separation

Fig.2 Binary image of rice scanning image and image comparison before and after impurity removal

Fig.3 Rice image in black background

Fig.4 Minimum circumscribed rectangle image of rice image

Fig.5 Contour image of rice leaves and references

Fig.6 Topological structure of BP neural network model

Fig.7 Analysis table of rice leaf length and leaf width error

Fig.8 Correlation coefficient distribution of rice image characteristics and leaf total nitrogen content Feature numbers 1 to 9 in the figure represented leaf R, leaf G, leaf B, leaf NRI, leaf NGI, leaf NBI, leaf H, leaf S, leaf I, respectively. Feature numbers 10 to 18 represented leaf sheath R, leaf sheath G, leaf sheath B, leaf sheath NRI, leaf sheath NGI, leaf sheath NBI, leaf sheath H, leaf sheath S, leaf sheath I, respectively. Feature numbers 19 to 25 represented leaf length, leaf width, leaf perimeter, leaf area, area perimeter ratio, area length ratio, length width ratio, respectively. The red values showed bilateral significant correlation at 0.010 level; the blue values showed bilateral significant correlation at 0.050 level.

Table 1 Multivariate linear regression modeling independent variable selection and model test results

时期及叶位 Period and leaf position	选择特征 Selected characteristics	多元线性回归模型 Multivariate regression model	$R c 2$	$R v 2$	MRE_v	RMSE_v
幼穗分化期顶三叶 3rd leaf of the Young panicle stage	叶片R、叶片G、叶片NRI、叶片H、叶片宽度 Leaf R, leaf G, leaf NRI, leaf H, leaf width	y=5.800-0.008x₁-0.025x₂- 4.008x₃+0.010x₄+0.128x₅	0.862	0.875	0.035	0.094
齐穗期顶二叶 2nd leaf of the full heading stage	叶片NRI、叶片NBI、叶片H、叶鞘NRI、叶鞘H Leaf NRI, leaf NBI, leaf H, leaf sheath NRI, leaf sheath H	y=-7.917-0.519x₁-1.341x₂+ 0.118x₃+1.804x₄-0.002x₅	0.708	0.714	0.058	0.164
齐穗期顶三叶 3rd leaf of the full heading stage	叶鞘B、叶鞘NGI、叶鞘NBI、叶鞘S、叶鞘I Leaf sheath B, leaf sheath NGI, leaf sheath NBI, Leaf sheath S, leaf sheath I	y=2.743-0.043x₁-0.011x₂+ 15.004x₃-1.175x₄-2.380x₅	0.692	0.676	0.145	0.245

Table 1 Multivariate linear regression modeling independent variable selection and model test results

时期及叶位 Period and leaf position	选择特征 Selected characteristics	多元线性回归模型 Multivariate regression model	$R c 2$	$R v 2$	MRE_v	RMSE_v
幼穗分化期顶三叶 3rd leaf of the Young panicle stage	叶片R、叶片G、叶片NRI、叶片H、叶片宽度 Leaf R, leaf G, leaf NRI, leaf H, leaf width	y=5.800-0.008x₁-0.025x₂- 4.008x₃+0.010x₄+0.128x₅	0.862	0.875	0.035	0.094
齐穗期顶二叶 2nd leaf of the full heading stage	叶片NRI、叶片NBI、叶片H、叶鞘NRI、叶鞘H Leaf NRI, leaf NBI, leaf H, leaf sheath NRI, leaf sheath H	y=-7.917-0.519x₁-1.341x₂+ 0.118x₃+1.804x₄-0.002x₅	0.708	0.714	0.058	0.164
齐穗期顶三叶 3rd leaf of the full heading stage	叶鞘B、叶鞘NGI、叶鞘NBI、叶鞘S、叶鞘I Leaf sheath B, leaf sheath NGI, leaf sheath NBI, Leaf sheath S, leaf sheath I	y=2.743-0.043x₁-0.011x₂+ 15.004x₃-1.175x₄-2.380x₅	0.692	0.676	0.145	0.245

Table 2 Comparison of validation results of estimation model of total nitrogen content in sensitive leaves of rice

模型 Model	幼穗分化期顶三叶 3rd leaf at the young panicle stage			齐穗期顶二叶 2nd leaf at the full heading stage
模型 Model	$R v 2$	MRE_v	RMSE_v	$R v 2$	MRE_v	RMSE_v
支持向量机Support vector machine	0.886	0.034	0.090	0.785	0.053	0.142
BP神经网络BP neural network	0.887	0.034	0.089	0.820	0.046	0.132

Table 2 Comparison of validation results of estimation model of total nitrogen content in sensitive leaves of rice

模型 Model	幼穗分化期顶三叶 3rd leaf at the young panicle stage			齐穗期顶二叶 2nd leaf at the full heading stage
模型 Model	$R v 2$	MRE_v	RMSE_v	$R v 2$	MRE_v	RMSE_v
支持向量机Support vector machine	0.886	0.034	0.090	0.785	0.053	0.142
BP神经网络BP neural network	0.887	0.034	0.089	0.820	0.046	0.132

Fig.9 Fitting results of actual value and predicted value of 3rd leaf at the young panicle stage based on machine learning a, Support vector machine model; b, BP neural network model.Same as Figure 10.

Fig.10 Fitting results of actual value and predicted value of 2nd leaf at the full heading stage based on machine learning

References 30

[1]	YU X J, LU H D, LIU Q Y. Deep-learning-based regression model and hyperspectral imaging for rapid detection of nitrogen concentration in oilseed rape (Brassica napus L.) leaf[J]. Chemometrics and Intelligent Laboratory Systems, 2018,172:188-193.
[2]	CAMERON K C, DI H J, MOIR J L. Nitrogen losses from the soil/plant system: a review[J]. Annals of Applied Biology, 2013,162(2):145-173.
[3]	LEE K J, LEE B W. Estimation of rice growth and nitrogen nutrition status using color digital camera image analysis[J]. European Journal of Agronomy, 2013,48:57-65.
[4]	李岚涛, 张萌, 任涛, 等. 应用数字图像技术进行水稻氮素营养诊断[J]. 植物营养与肥料学报, 2015,21(1):259-268.
	LI L T, ZHANG M, REN T, et al. Diagnosis of N nutrition of rice using digital image processing technique[J]. Journal of Plant Nutrition and Fertilizer, 2015,21(1):259-268.(in Chinese with English abstract)
[5]	YUAN Y, CHEN L, LI M, et al. Diagnosis of nitrogen nutrition of rice based on image processing of visible light[C]//. IEEE, 2016: 228-232.
[6]	顾清, 邓劲松, 陆超, 等. 基于光谱和形状特征的水稻扫描叶片氮素营养诊断[J]. 农业机械学报, 2012,43(8):170-174.
	GU Q, DENG J S, LU C, et al. Diagnosis of rice nitrogen nutrition based on spectral and shape characteristics of scanning leaves[J]. Transactions of the Chinese Society for Agricultural Machinery, 2012,43(8):170-174.(in Chinese with English abstract)
[7]	周琼, 杨红云, 杨珺, 等. 基于BP神经网络和概率神经网络的水稻图像氮素营养诊断[J]. 植物营养与肥料学报, 2019,25(1):134-141.
	ZHOU Q, YANG H Y, YANG J, et al. Feasibility study of BP neural network and probabilistic neural network for nitrogen nutrition diagnosis of rice images[J]. Journal of Plant Nutrition and Fertilizers, 2019,25(1):134-141.(in Chinese with English abstract)
[8]	罗正媛, 汤洋. 黑龙江省玉米产量变化的预测分析: 基于支持向量机的实证研究[J]. 农机化研究, 2013,35(2):30-34.
	LUO Z Y, TANG Y. The prediction analysis on maize yield changes in Heilongjiang Province: empirical research based on support vector machine(SVM)[J]. Journal of Agricultural Mechanization Research, 2013,35(2):30-34.(in Chinese with English abstract)
[9]	张建华, 冀荣华, 袁雪, 等. 基于径向基支持向量机的棉花虫害识别[J]. 农业机械学报, 2011,42(8):178-183.
	ZHANG J H, JI R H, YUAN X, et al. Recognition of pest damage for cotton leaf based on RBF-SVM algorithm[J]. Transactions of the Chinese Society for Agricultural Machinery, 2011,42(8):178-183.(in Chinese with English abstract)
[10]	朱文静, 毛罕平, 周莹, 等. 基于高光谱图像技术的番茄叶片氮素营养诊断[J]. 江苏大学学报(自然科学版), 2014,35(3):290-294.
	ZHU W J, MAO H P, ZHOU Y, et al. Hyperspectral imaging technology of nitrogen status diagnose for tomato leaves[J]. Journal of Jiangsu University (Natural Science Edition), 2014,35(3):290-294.(in Chinese with English abstract)
[11]	陈利苏. 基于机器视觉技术的水稻氮磷钾营养识别和诊断[D]. 杭州: 浙江大学, 2014.
	CHEN L S. Rice nutrition identification and diagnosis based on machine vision technology[D]. Hangzhou: Zhejiang University, 2014.(in Chinese with English abstract)
[12]	彭莹琼, 廖牧鑫, 张永红, 等. 基于BP神经网络模型的果实蝇自动分类系统[J]. 江西农业大学学报, 2016,38(6):1205-1210.
	PENG Y Q, LIAO M X, ZHANG Y H, et al. A study on the automatic classification system for fruit flies based on BP neural network model[J]. Acta Agriculturae Universitatis Jiangxiensis, 2016,38(6):1205-1210.(in Chinese with English abstract)
[13]	潘圣刚, 黄胜奇, 张帆, 等. 超高产栽培杂交中籼稻的生长发育特性[J]. 作物学报, 2011,37(3):537-544.
	PAN S G, HUANG S Q, ZHANG F, et al. Growth and development characteristics of super-high-yielding mid-season indica hybrid rice[J]. Acta Agronomica Sinica, 2011,37(3):537-544.(in Chinese with English abstract) DOI URL
[14]	孙小香, 王芳东, 赵小敏, 等. 基于冠层光谱和BP神经网络的水稻叶片氮素浓度估算模型[J]. 中国农业资源与区划, 2019,40(3):35-44.
	SUN X X, WANG F D, ZHAO X M, et al. The estimation models of rice leaf nitrogen concentration based on canopy spectrum and BP neural network[J]. Chinese Journal of Agricultural Resources and Regional Planning, 2019,40(3):35-44.(in Chinese with English abstract)
[15]	程慧煌, 商庆银, 易振波, 等. 不同产量水平超级杂交稻产量形成特征及其对施肥量的响应[J]. 中国稻米, 2017,23(4):81-88.
	CHENG H H, SHANG Q Y, YI Z B, et al. Effects of fertilizer application rate on yield and population quality of super hybrid rice at different yield levels[J]. China Rice, 2017,23(4):81-88.(in Chinese with English abstract)
[16]	林洪鑫, 潘晓华, 石庆华, 等. 栽插密度与施氮量对双季稻上部三叶叶长和叶角的影响[J]. 作物学报, 2010,36(10):1743-1751.
	LIN H X, PAN X H, SHI Q H, et al. Effects of nitrogen application amount and planting density on angle and length of top three leaves in double-cropping rice[J]. Acta Agronomica Sinica, 2010,36(10):1743-1751.(in Chinese with English abstract)
[17]	孙玉婷, 王映龙, 杨红云, 等. RGB与HSI色彩空间下预测叶绿素相对含量的研究[J]. 浙江农业学报, 2018,30(10):1782-1789.
	SUN Y T, WANG Y L, YANG H Y, et al. Prediction of SPAD in rice leaf based on RGB and HSI color space[J]. Acta Agriculturae Zhejiangensis, 2018,30(10):1782-1789.(in Chinese with English abstract)
[18]	杨红云, 孙爱珍, 何火娇. 水稻叶片几何参数图像视觉测量方法研究[J]. 湖北农业科学, 2015,54(17):4317-4320.
	YANG H Y, SUN A Z, HE H J. Study on the geometry parameter of rice leaf measuring method using image vision technology[J]. Hubei Agricultural Sciences, 2015,54(17):4317-4320.(in Chinese with English abstract)
[19]	崔世钢, 秦建华, 张永立. 基于图像处理技术的植物叶片面积和周长测量[J]. 江苏农业科学, 2018,46(15):187-189.
	CUI S G, QIN J H, ZHANG Y L. Measurement of leaf area and girth based on image processing technology[J]. Jiangsu Agricultural Sciences, 2018,46(15):187-189.(in Chinese with English abstract)
[20]	吴大伟, 李聪, 徐丽, 等. 基于图像处理技术的植物叶片相关参数测量研究[J].科技信息, 2010(4):106.
	WU D W, LI C, XU L, et al. Measurement of plant leaf parameters based on image processing technology[J]. Science & Technology Information, 2010(4):106.
[21]	李朝东, 崔国贤, 谢宁, 等. 应用数字图像技术诊断苎麻氮素营养的研究简报[J]. 植物营养与肥料学报, 2011,17(3):767-772.
	LI C D, CUI G X, XIE N, et al. Research notes on N status diagnosis of ramie by using digital image technology[J]. Plant Nutrition and Fertilizer Science, 2011,17(3):767-772.(in Chinese with English abstract)
[22]	孙铁波, 王卫兵, 刘春月. 嵌入式水稻氮素营养检测仪的研究与设计[J]. 中国农机化学报, 2015,36(3):257-261.
	SUN T B, WANG W B, LIU C Y. Research and design of the embedded detector for rice nitrogen nutrition[J]. Journal of Chinese Agricultural Mechanization, 2015,36(3):257-261.(in Chinese with English abstract)
[23]	谭昌伟, 周清波, 齐腊, 等. 水稻氮素营养高光谱遥感诊断模型[J]. 应用生态学报, 2008,19(6):1261-1268. PMID
	TAN C W, ZHOU Q B, QI L, et al. Hyperspectral remote sensing diagnosis models of rice plant nitrogen nutritional status[J]. Chinese Journal of Applied Ecology, 2008,19(6):1261-1268.(in Chinese with English abstract) URL PMID
[24]	高创, 郁崇文, 汪军, 等. 棉花性能指标对成纱质量的预测模型研究[J]. 棉纺织技术, 2016,44(6):18-22.
	GAO C, YU C W, WANG J, et al. Prediction model study of cotton quality index on yarn quality[J]. Cotton Textile Technology, 2016,44(6):18-22.(in Chinese with English abstract)
[25]	常军, 李祯, 朱业玉, 等. 基于支持向量机(SVM)方法的冬季温度预测[J]. 气象科技, 2005,33(增刊):100-104.
	CHANG J, LI Z, ZHU Y Y, et al. Application of support vector machine method to winter temperature forecast[J]. Scientia Meteorological Sinica, 2005,33(Suppl.):100-104.(in Chinese with English abstract)
[26]	奉国和. SVM分类核函数及参数选择比较[J]. 计算机工程与应用, 2011,47(3):123-124.
	FENG G H. Parameter optimizing for support vector machines classification[J]. Computer Engineering and Applications, 2011,47(3):123-124.(in Chinese with English abstract)
[27]	周琼, 杨红云, 杨珺, 等. 基于参数优化支持向量机的水稻施氮水平分类研究[J]. 南方农业学报, 2017,48(8):1524-1528.
	ZHOU Q, YANG H Y, YANG J, et al. Classification of nitrogen application level for rice based on support vector machine optimized by parameters[J]. Journal of Southern Agriculture, 2017,48(8):1524-1528.(in Chinese with English abstract)
[28]	胡强, 郝晓燕, 雷蕾. 基于遗传算法和BP神经网络的孤立性肺结节分类算法[J]. 计算机科学, 2016,43(6A):37-39.
	HU Q, HAO X Y, LEI L. Solitary pulmonary nodules classification based on genetic algorithm and back propagation neural networks[J]. Computer Science, 2016,43(6A):37-39.(in Chinese with English abstract)
[29]	师黎, 朱民杰. 基于遗传算法优化BP神经网络在心电图身份识别中的应用[J]. 中国组织工程研究与临床康复, 2010,14(43):8069-8072.
	SHI L, ZHU M J. Human identification using electrocardiograms based on generic algorithm-back propagation neural network[J]. Journal of Clinical Rehabilitative Tissue Engineering Research, 2010,14(43):8069-8072.(in Chinese with English abstract)
[30]	郭士伟, 赵学强, 夏士健, 等. 超级杂交稻生育后期叶片和根系的衰老营养生理研究[J]. 华北农学报, 2014,29(3):115-121.
	GUO S W, ZHAO X Q, XIA S J, et al. The leaf and root nourishment physiology research for the super-hybrid rice after heading[J]. Acta Agriculturae Boreali-Sinica, 2014,29(3):115-121.(in Chinese with English abstract)

Study on estimation model of total nitrogen content in rice leaves based on image characteristics

RichHTML

PDF (PC)

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 12

References 30

Related Articles 15

Recommended Articles

Metrics

Comments

[1]	CHEN Dan, TANG Cuifeng, DONG Chao, GAN Shuxian, LI Jun, A Xinxiang, ZHANG Feifei, YANG Yayun, NIU Saisai, DAI Luyuan. Grain starch quality characteristics of Yunnan soft rice landraces [J]. Acta Agriculturae Zhejiangensis, 2021, 33(2): 203-214.
[2]	ZOU Wenxiong, WU Wei, GUAN Yajing, CAO Dongdong, BIAN Xiaobo, SHI Deyun, DING Liling. Research progress of regulation techniques of rice seed dormancy [J]. Acta Agriculturae Zhejiangensis, 2021, 33(2): 369-379.
[3]	ZHENG Xusong, ZHONG Liequan, WANG Huifu, CHEN Fangjing, CHENG Liping, XU Qiqiang, LI Yang, ZHONG Xuhua, LYU Zhongxian. Demonstration on rice pests control by fertilizer regulation technique in different geographic rice growing areas of Zhejiang Province [J]. , 2020, 32(9): 1656-1664.
[4]	LIU Jiaojiao, ZHANG Gangren, GAO Qun. Study on price relevance of livestock products based on VEC-BEKK-GARCH and spillover index model [J]. , 2020, 32(9): 1711-1721.
[5]	LU Muyuan, LIU Yuan, LIU Guijian. Predictive analysis of soil organic matter content in black soil region based on combined model [J]. , 2020, 32(8): 1427-1436.
[6]	GU Jiancheng, GUO Bin, LIN Yicheng, FU Qinglin, LIU Chen, DING Nengfei, LI Hua, LI Ningyu. Effects of iron plaque in root surface on cadmium uptake of rice [J]. , 2020, 32(6): 963-970.
[7]	SHAO Wenqi, ZHONG Ping, DONG Yubing, SUN Chunmei, JI Li, ZHUANG Chun, CHEN Chuan, ZHANG Ankang. Difference of light and temperature resources in tray seedling and its effect on seedling quality in rice [J]. , 2020, 32(2): 191-199.
[8]	ZHANG Hongtao, LI Yijia, TAN Lian, XU Shuaitao. Image recognition of millet leaf disease based on CS-SVM [J]. , 2020, 32(2): 274-282.
[9]	WANG Nianyi, YU Fenghua, XU Tongyu, DU Wen, GUO Zhonghui, ZHANG Guosheng. Hyperspectral retrieval modelling for chlorophyll contents of japonica-rice leaves based on machine learning [J]. , 2020, 32(2): 359-366.
[10]	GAO Yun, ZHAO Zhijun. Influence of imported honey on market of bee products of China [J]. , 2020, 32(11): 2088-2093.
[11]	JIANG Junfang, LIU Junchao, WU Jianliang, ZHENG Huichao, HUANG Xin, ZHENG Kaizhi, JIANG Yongqing. Study on dynamic change of fermentation quality in mixed silages of bamboo shoot shell and rice husk [J]. , 2020, 32(10): 1757-1763.
[12]	FANG Yuwei, ZHANG Wei, CHEN Youyuan, HOU Fan, XU Linfeng, TANG Changhua, LI Rongde. State quo of utilization of high-quality hybrid rice varieties in China during 2001-2017 [J]. , 2020, 32(1): 1-14.
[13]	ZHANG Xinyue, LI Youfa, LIU Jiangning, FU Haowei. Study on hybrid rice purity of special combining pattern testing with functional marker of wide compatibility gene S5-n [J]. , 2020, 32(1): 15-19.
[14]	BING Jingjing, GAO Hongmei. Research on development of high quality rice industry in Tianjin based on SWOT analysis: Taking "Xiaozhan" rice as an example [J]. , 2019, 31(8): 1217-1223.
[15]	ZHANG Shun, YANG Jitao, LI Yong, LIAO Juan, LI Zhaodong, ZHU Dequan. Experiment of sucking performance of inside-filling pneumatic type precision hill-drop drilling seed-metering device for rice [J]. , 2019, 31(8): 1379-1387.