基于卷积神经网络的水稻氮素营养诊断

doi:10.13304/j.nykjdb.2022.0700

中国农业科技导报 ›› 2023, Vol. 25 ›› Issue (9): 113-121.DOI: 10.13304/j.nykjdb.2022.0700

基于卷积神经网络的水稻氮素营养诊断

钱政¹(), 杨孙哲², 张国卿³, 郭紫微⁴, 张林朋¹, 万家兴¹, 杨红云¹()

^1.江西农业大学软件学院，南昌 330045
^2.江西师范大学计算机信息工程学院，南昌 330022
^3.北京粉笔天下教育科技有限公司石家庄分公司，石家庄 050051
^4.江西农业大学计算机与信息工程学院，南昌 330045

收稿日期:2022-08-23 接受日期:2022-12-29 出版日期:2023-09-15 发布日期:2023-09-28
通讯作者: 杨红云
作者简介:钱政 E-mail：qz1058137670@outlook.com；
基金资助:
国家自然科学基金项目(62162030);江西省研究生创新专项资金项目(YC2022-s432)

Rice Nitrogen Nutrition Diagnosis Based on Convolutional Neural Network

Zheng QIAN¹(), Sunzhe YANG², Guoqing ZHANG³, Ziwei GUO⁴, Linpeng ZHANG¹, Jiaxing WAN¹, Hongyun YANG¹()

^1.School of Software，Jiangxi Agricultural University，Nanchang 330045，China
^2.School of Information Engineering，Jiangxi Normal University，Nanchang 330022，China
^3.Fenbi Technology Ltd. Shijiazhuang Branch，Fenbi Technology Co. ，Ltd. ，Shijiazhuang 050051，China
^4.School of Computer and Information Engineering，Jiangxi Agricultural University，Nanchang 330045，China

Received:2022-08-23 Accepted:2022-12-29 Online:2023-09-15 Published:2023-09-28
Contact: Hongyun YANG

摘要/Abstract

摘要：

为了快速、准确诊断和识别水稻氮素胁迫程度，对水稻进行大田栽培试验。以超级水稻‘两优培九’为试验对象，设置0、210、300和390 kg·hm^-2共4个施氮水平处理，通过扫描采集幼穗分化期和齐穗期水稻顶1、顶2、顶3叶图像，在卷积神经网络（convolutional neural network，CNN）ResNet34的每个残差块中加入SE block（squeeze-and-excitation block）模块，并将在图像数据集ImageNet（ImageNet large scale visual recognition challenge）上训练得到的权重参数迁移到水稻氮素营养诊断的识别模型中，ResNet34的特征提取层保持原结构，模型结尾的池化层替换为全局平均池化层，利用改进后的网络对水稻图像进行特征提取，训练得到最优的权重参数。结果表明，改进后的网络对水稻幼穗分化期的模型测试准确率达到98.13%，齐穗期的准确率达到99.46%，且模型的收敛速度更快，相比于改进前的网络准确率均提升了7%以上。以上结果表明，通过在ResNet34残差块中加入SE block并基于迁移学习的方法对水稻氮素营养诊断方法是可行的，能有效对水稻幼穗分化期和齐穗期的氮素营养进行诊断识别，为农作物的营养诊断识别提供了参考。

关键词: 卷积神经网络, 水稻氮素营养诊断, ResNet34, 迁移学习, SE block

Abstract:

In order to rapidly and accurately diagnose and identify nitrogen stress in rice， a field experiment was conducted. Taking the super rice variety ‘Liangyoupei 9’as material， 4 treatments of nitrogen application （0， 210， 300， and 390 kg·hm^-2） were set， and images of the first， second， and third leaves at the top of the rice plant were scanned and collected during the spikelet differentiation and the full heading stage. The SE block （squeeze-and-excitation block） module was added to each residual block of the ResNet34 in convolutional neural network （CNN）， and the weight parameters trained on the ImageNet （ImageNet large scale visual recognition challenge） dataset were transferred to the nitrogen nutrition diagnosis model of rice. The feature extraction layer of ResNet34 was kept unchanged， and the pooling layer at the end of the model was replaced with a global average pooling layer. The improved network was used to extract features from rice images and train the optimal weight file. The results showed that the improved network achieved a testing accuracy of 98.13% during the spikelet differentiation stage and 99.46% during the full heading stage of rice. The convergence speed of the model was faster， and the accuracy was improved by more than 7% compared to the original network. Above results showed that it was feasible to add the SE block to the residual block of ResNet34 and use transfer learning to diagnose nitrogen nutrition in rice， which could effectively diagnose and identify the nitrogen nutrition of rice during the spikelet differentiation stage and the full heading stage. Above results provided reference for the diagnosis and identification of nutrient status in crops.

Key words: convolutional neural network, rice nitrogen nutrition diagnosis, ResNet34, transfer learning, SE block

中图分类号:

TP181

钱政, 杨孙哲, 张国卿, 郭紫微, 张林朋, 万家兴, 杨红云. 基于卷积神经网络的水稻氮素营养诊断[J]. 中国农业科技导报, 2023, 25(9): 113-121.

Zheng QIAN, Sunzhe YANG, Guoqing ZHANG, Ziwei GUO, Linpeng ZHANG, Jiaxing WAN, Hongyun YANG. Rice Nitrogen Nutrition Diagnosis Based on Convolutional Neural Network[J]. Journal of Agricultural Science and Technology, 2023, 25(9): 113-121.

图/表 6

表1 原始数据集数量

Table 1 Original rice data number

施氮处理 Nitrogen application treatment	原始图像数量 Number of original images
施氮处理 Nitrogen application treatment	幼穗分化期 Spikelet differentiation stage	齐穗期 Full heading stage
N1	240	232
N2	225	233
N3	240	235
N4	216	225

图1 改进的模型网络构架

Fig. 1 Improved model network architecture

表2 不同方案的试验数据

Table 2 Experimental data for different plans

时期 Stage	方案 Plan	训练时间 Train time/s	参数量 Parameter	准确率 Accuracy/%	召回率 Recall/%
幼穗分化期 Spikelet differentiation stage	1	6 438.434	21 808 964	91.68	91.74
	2	6 926.372	21 447 920	93.55	94.09
	3	7 084.370	21 447 920	98.13	98.15
齐穗期 Full heading stage	1	6 519.452	21 808 964	92.02	92.09
	2	7 020.321	21 447 920	94.15	94.16
	3	7 103.854	21 447 920	99.46	99.47

图2 3个方案在幼穗分化期的训练曲线和混淆矩阵A：训练集损失曲线； B：测试集损失曲线； C：测试集准确率曲线； D：方案1的ResNet34混淆矩阵； E：方案2的ResNet34_SE混淆矩阵； F：方案3的ResNet34_SEt混淆矩阵

Fig. 2 Training curves and confusion matrices of three schemes during the spikelet differentiation stageA：Curve of train loss； B：Curve of test loss； C：Curve of test accuracy； D： ResNet34 confusion matrix of plan 1； E： ResNet34_SE confusion matrix of plan 2； F： ResNet34_SEt confusion matrix of plan 3

图3 3个方案在齐穗期的训练曲线和混淆矩阵A：训练集损失曲线； B：测试集损失曲线； C：测试集准确率曲线； D：方案1的ResNet34混淆矩阵； E：方案2的ResNet34_SE混淆矩阵； F：方案3的ResNet34_SEt混淆矩阵

Fig. 3 Training curves and confusion matrices of three schemes during the full heading stageA：Curve of train loss； B：Curve of test loss； C：Curve of test accuracy； D： ResNet34 confusion matrix of plan 1； E： ResNet34_SE confusion matrix of plan 2； F： ResNet34_SEt confusion matrix of plan 3

表3 不同模型的性能对比

Table 3 Performance comparison of different models

模型 Model	时期 Stage	准确率 Accuracy/%	训练时间 Train time/s	参数量 Parameters number
AlexNet	幼穗分化期 Spikelet differentiation stage	66.67	7 255.538	255 201 092
AlexNet	齐穗期 Full heading stage	92.15	7 514.199	255 201 092
VGG11	幼穗分化期 Spikelet differentiation stage	93.43	16 703.601	562 893 188
VGG11	齐穗期 Full heading stage	93.70	15 300.326	562 893 188
VGG16	幼穗分化期 Spikelet differentiation stage	84.24	28 812.880	568 387 396
VGG16	齐穗期 Full heading stage	94.93	28 917.401	568 387 396
ResNet34_SEt	幼穗分化期 Spikelet differentiation stage	98.13	7 084.370	21 808 964
ResNet34_SEt	齐穗期 Full heading stage	99.46	7 103.854	21 808 964

参考文献 28

1	朱永波,韩展誉,程方民. 氮肥对水稻营养品质、外观品质和加工品质的影响[J]. 基层农技推广, 2019,7(12):26-29.
2	YU X, LU H, LIU Q. Deep learning based regression model and hyperspectral imaging for rapid detection of nitrogen concentration in oilseed rape (Brassica napus L.) leaf [J]. Chemom. Intell. Lab. Syst., 2018,172:188-193.
3	CAMERON K C, DI H J, MOIR J L. Nitrogen losses from the soil/plant system: a review [J]. Ann. Appl. Biol., 2013,162(2):145-173.
4	刘晓伟,王火焰,朱德进,等. 氮肥施用方式对水稻产量以及氮、磷、钾养分吸收利用的影响[J]. 南京农业大学学报, 2017,40(2):203-210.
	LIU X W, WANG H Y, ZHU D J, et al.. Effect of N fertilization method on rice yield and N, P and K uptake and use efficiency [J]. J. Nanjing Agric. Univ., 2017,40(2):203-210.
5	孙棋. 基于数字图像处理技术的水稻氮素营养诊断研究[D]. 杭州:浙江大学,2008.
	SUN Q. Rice nitrogen nutrition diagnosis based on digital image processing technique [D]. Hangzhou:Zhejiang University, 2008.
6	张国圣,许童羽,于丰华,等. 基于高光谱的水稻叶片氮素估测与反演模型[J]. 浙江农业学报, 2017,29(5):845-849.
	ZHANG G S, XU T Y, YU F H, et al.. Nitrogen content inversion of rice leaf based on the hyperspectral data [J]. Zhejiang Agric. J., 2017,29(5):845-849.
7	HUANG S, MIAO Y, ZHAO G, et al.. Estimating rice nitrogen status with satellite remote sensing in Northeast China [C]//Proceedings of 2013 Second International Conference on Agro-Geoinformatics (Agro-Geoinformatics). USA Virginia, IEEE, 2013:550-557.
8	郭建华,王秀,孟志军,等. 主动遥感光谱仪Greenseeker与SPAD对玉米氮素营养诊断的研究[J]. 植物营养与肥料学报, 2008, 14(1):43-47.
	GUO J H, WANG X, MENG Z J, et al.. Study on diagnosing nitrogen nutrition status of corn using Greenseeker and SPAD meter [J]. J. Plant Nutr. Fert., 2008,14(1):43-47.
9	杨红云,罗建军,万颖,等. 计算机视觉技术在水稻氮素营养诊断中应用的研究进展[J]. 中国农学通报, 2020,36(16):149-155.
	YANG H Y, LUO J J, WAN Y, et al.. Application of computer vision technology in nitrogen nutrition diagnosis of rice: research progress [J]. Chin. Agric. Sci. Bull., 2020,36(16):149-155.
10	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]. J. Plant Nutr. Fert., 2015,21(1):259 -268.
11	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]. J. Plant Nutr. Fert., 2019,25(1):134-141.
12	XU Z, GUO X, ZHU A, et al.. Using deep convolutional neural networks for image-based diagnosis of nutrient deficiencies in rice [J/OL]. Comput. Intell. Neurosci., 2020,9:7307252 [2022-07-22]..
13	YUAN Y, CHEN L, LI M, et al.. Diagnosis of nitrogen nutrition of rice based on image processing of visible light [C]// Proceedings of 2016 IEEE International Conference on Functional-Structural Plant Growth Modeling, Simulation, Visualization and Applications (FSPMA). China Qingdao, IEEE, 2016:228-232.
14	SATHYAVANI R, JAGANMOHAN K, KALAAVATHI B. Classification of nutrient deficiencies in rice crop using denseNet-BC [J]. Mater. Today: Proceed., 2022,56:1783-1789.
15	WANG C, YE Y, TIAN Y, et al.. Classification of nutrient deficiency in rice based on CNN model with reinforcement learning augmentation [C]// Proceedings of 2021 International Symposium on Artificial Intelligence and its Application on Media (ISAIAM). China Xi’an, IEEE, 2021:107-111.
16	HU J, SHEN L, SUN G. Squeeze-and-excitation networks [C]//Proceedings of the IEEE conference on computer vision and pattern recognition. USA Salt Lake City, IEEE, 2018:7132-7141.
17	杨红云,肖小梅,黄琼,等. 基于卷积神经网络和迁移学习的水稻害虫识别[J]. 激光与光电子学进展,2022,59(16): 333-340.
	YANG H Y, XIAO X M, HUANG Q, et al.. Rice pest identification based on convolutional neural network and transfer learning [J]. Adv. Lasers Optoelectr., 2022,59(16): 333-340.
18	姚强,粟超,李波,等. 深度学习方法在水稻氮素营养诊断中的应用初探[J]. 南方农业,2021,15(31):125-129.
19	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.. Feasibility study of BP neural network and probabilistic neural network for nitrogen nutrition diagnosis of rice images [J]. China Rice, 2017,23(4):81-88.
20	孙振宇,黄巍,于艳丽,等. 提高水稻区试质量的主要技术措施[J]. 吉林农业,2009(2):20.
21	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 Agron. Sin., 2010,36(10):1743-1751.
22	RICE L, WONG E, KOLTER Z. Overfitting in adversarially robust deep learning [C]//International Conference on Machine Learning. Austria Vienna, PMLR, 2020:8093-8104.
23	SHORTEN C, KHOSHGOFTAAR T M. A survey on image data augmentation for deep learning [J/OL]. J. Big Data, 2019,6(1):197[2022-07-22]. .
24	张顺,龚怡宏,王进军. 深度卷积神经网络的发展及其在计算机视觉领域的应用[J]. 计算机学报,2019,42(3):453-482.
	ZHANG S, GONG Y H, WANG J J. The development of deep convolution neural network and its applications on computer vision [J]. Chin. J. Comput., 2019,42(3):453-482.
25	HE K, ZHANG X, REN S, et al.. Deep residual learning for image recognition [C]// Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. USA Las Vegas, IEEE, 2016:770-778.
26	罗建军,杨红云,路艳,等. 基于遗传算法优化的BP神经网络进行水稻氮素营养诊断[J]. 中国农业科技导报,2020,22(8):83-92.
	LUO J J, YANG H Y, LU Y, et al.. Identification of nitrogen nutrition in rice based on BP neural network optimized by genetic algorithms [J]. J. Agric. Sci. Technol., 2020,22(8):83-92.
27	张林朋,杨红云,钱政,等. 基于改进的VGG16网络和迁移学习的水稻氮素营养诊断[J].中国农业大学学报,2023,28(6):219-229.
	ZHANG L P, YANG H Y, QIAN Z, et al.. Nitrogen nutrition diagnosis in rice based on improved VGG16 network and transfer learning [J]. J. Chin. Agric. Univ., 2023,28(6):219-229.
28	CAP Q H, UGA H, KAGIWADA S, et al.. LeafGAN: an effective data augmentation method for practical plant disease diagnosis [J]. IEEE Trans. Autom. Sci. Eng., 2022,19(2):1258-1267.

[1]	林开颜, 梅飞, 吴军辉, 郭文刚, 陈杰, 司慧萍. 基于计算机视觉的作物病害监测服务平台设计与研究[J]. 中国农业科技导报, 2023, 25(6): 89-96.
[2]	赵威, 马睿, 王佳, 郭宏杰, 许金普. 基于果穗图像的玉米品种分类识别[J]. 中国农业科技导报, 2023, 25(6): 97-106.
[3]	方逵, 李成, 何潇, 陈益能. 基于三维重建的多角度葡萄叶病害识别方法研究[J]. 中国农业科技导报, 2022, 24(7): 86-96.
[4]	黄昊, 谢圣桥, 陈度, 王恒. 深度学习在苹果产业链中的应用与研究进展[J]. 中国农业科技导报, 2022, 24(10): 79-89.
[5]	周惠汝, 吴波明. 深度学习在作物病害图像识别方面应用的研究进展[J]. 中国农业科技导报, 2021, 23(5): 61-68.
[6]	郭阳, 许贝贝, 陈桂鹏, 丁建, 严志雁, 梁华, 吴昌华. 基于卷积神经网络的水稻虫害识别方法[J]. 中国农业科技导报, 2021, 23(11): 99-109.
[7]	董伟1,钱蓉1,张洁2*,张立平1,陈红波2,张萌1,朱静波1,卜英乔3. 基于深度学习的蔬菜鳞翅目害虫自动识别与检测计数[J]. 中国农业科技导报, 2019, 21(12): 76-84.
[8]	刘瑾蓉,林剑辉*,李婷婷. 基于卷积神经网络的银杏叶片患病程度识别[J]. 中国农业科技导报, 2018, 20(6): 55-61.

基于卷积神经网络的水稻氮素营养诊断

Rice Nitrogen Nutrition Diagnosis Based on Convolutional Neural Network

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 6

参考文献 28

相关文章 8

编辑推荐

Metrics