基于深度学习的香蕉病害远程诊断系统

史红栩; 李修华; 李民赞; 王伟; 温标堂

doi:10.7671/j.issn.1001-411X.202004027

基于深度学习的香蕉病害远程诊断系统

史红栩^1,,
李修华^{1, 2, ,},
李民赞³,
王伟¹,
温标堂⁴

1.
广西大学电气工程学院，广西南宁 530004
2.
广西甘蔗生物学重点实验室，广西南宁 530004
3.
中国农业大学现代精细农业系统集成研究教育部重点实验室，北京 100083
4.
广西捷佳润科技有限公司，广西南宁 530001

基金项目: 国家自然科学基金(31760342)；广西科技重大专项经费项目(桂科AA18118037)

详细信息

作者简介:
史红栩(1994—)，男，硕士研究生，E-mail: 523256095@qq.com

通讯作者:
李修华(1983—)，女，副教授，博士，E-mail: lixh@gxu.edu.cn

中图分类号: S126
计量
- 文章访问数: 1081
- HTML全文浏览量: 8
- PDF下载量: 1266
出版历程
- 收稿日期: 2020-04-20
- 网络出版日期: 2023-05-17
- 刊出日期: 2020-11-09

Remote diagnosis system of banana diseases based on deep learning

SHI Hongxu^1,,
LI Xiuhua^{1, 2, ,},
LI Minzan³,
WANG Wei¹,
WEN Biaotang⁴

1.
School of Electrical Engineering, Guangxi University, Nanning 530004, China
2.
Guangxi Key Laboratory for Sugarcane Biology, Nanning 530004, China
3.
Key Laboratory of Modern Precision Agriculture System Integration Research, Ministry of Education, China Agricultural University, Beijing 100083, China
4.
Guangxi Jie Jia Run Technology Co., Ltd., Nanning 530001, China

摘要

摘要:
目的
实现香蕉病害的远程诊断。

方法
基于深度学习方法对香蕉作物的7种常见病害进行诊断。收集了5944幅健康及染病香蕉植株图像，按7∶1∶2分为训练集、验证集和测试集。利用迁移学习对GoogLeNet深度卷积神经网络训练获取诊断模型。进一步开发了包含手机移动应用程序(APP)和远程服务器的软件系统。

结果
通过对比不同迭代次数及不同优化器，最终采用了MomentumOptimizer迭代10000次的模型，平均测试精度达到了98%。设计的APP能够就地获取香蕉图像，并通过网络与集成了诊断模型的远程服务器通信，实时获取诊断结果。

结论
该病害诊断模型识别主要病害的精度高，在线诊断系统简单易操作，可快速有效地在线诊断香蕉常见病害，具有良好的应用前景。
- 香蕉病害 /
- 深度学习 /
- GoogLeNet /
- 图像识别 /
- 移动应用 /
- 迁移学习
Abstract:
Objective
To realize remote diagnosis of banana diseases.

Method
Deep learning method was used to diagnose seven common diseases of banana plant. A total of 5 944 images of diseased and healthy banana plants were collected and divided into training set, validation set and testing set according to the ratio of 7∶1∶2. Transfer learning was used to train GoogLeNet which is a deep convolutional neural network for obtaining the diagnosis model. A software system including a mobile application (APP) and a remote server was further developed.

Result
By comparing different iteration times and optimizers, the model of MomentumOptimizer with 10000 iteration times was finally selected, and the average test accuracy was 98%. The designed mobile APP could acquire banana images in situ, and communicate with the remote server which was integrated with a diagnosis model via the network to obtain diagnosis results in real time.

Conclusion
The disease diagnosis model can identify the main diseases with high accuracy. The online diagnosis system is simple and easy to operate, it can diagnose common banana diseases online quickly and effectively, and therefore it has a wide application prospect.
- banana disease /
- deep learning /
- GoogLeNet /
- image recognition /
- mobile application /
- transfer learning

HTML全文

图 1 香蕉病害诊断系统总体结构

Figure 1. Overall structure of the banana disease diagnosis system

下载: 全尺寸图片幻灯片

图 2 健康及染病的香蕉植株

Figure 2. Healthy and diseased banana plants

下载: 全尺寸图片幻灯片

图 3 Inception-V3结构

Figure 3. Inception-V3 structure

下载: 全尺寸图片幻灯片

图 4 不同优化器不同迭代次数下的模型训练结果

Figure 4. Model training results using different optimizers with different iteration times

下载: 全尺寸图片幻灯片

图 5 软件系统总体结构图

Figure 5. General structure of the software system

下载: 全尺寸图片幻灯片

图 6 APP测试结果

Figure 6. APP test result

下载: 全尺寸图片幻灯片

表 1 各类别样本图像数量

Table 1 Number of sample images in each category 幅

植株类别 Category	原始图像 Original image	数据增强 Data augmentation
健康 Healthy	130	1048
香蕉枯萎病 Fusarium wilt	217	1736
香蕉叶斑病 Banana leaf spot	170	1360
香蕉束顶病 Banana bunchy top	44	352
香蕉黑星病 Macrophoma musae	86	696
香蕉炭疽病 Banana anthracnose	38	304
香蕉黑疫病 Banana black blight	32	256
香蕉鞘腐病 Banana sheath rot	24	192
总计 Total	741	5944

下载: 导出CSV

表 2 模型的训练时间和平均测试精度

Table 2 Model training time and average test accuracy

迭代次数 Iteration	优化器 Optimizer	训练时间/min Training time	平均测试精度/% Average test accuracy
5000	GradientDescentOptimizer	11	94.5
10000	GradientDescentOptimizer	22	96.0
50000	GradientDescentOptimizer	110	98.0
10000	MomentumOptimizer	23	98.0
10000	AdamOptimizer	21	97.8

下载: 导出CSV

表 3 模型测试的混淆矩阵

Table 3 Confusion matrix of model testing result

类别 Category	健康 Healthy	香蕉枯萎病 Fusarium wilt	香蕉叶斑病 Banana leaf spot	香蕉束顶病 Banana bunchy top	香蕉黑星病 Macrophoma musae	香蕉炭疽病 Banana anthracnose	香蕉黑疫病 Banana black blight	香蕉鞘腐病 Banana sheath rot	总计 Total
健康 Healthy	210	0	0	1	0	0	0	0	211
香蕉枯萎病 Fusarium wilt	0	345	0	1	0	0	0	2	348
香蕉叶斑病 Banana leaf spot	0	0	263	0	3	1	0	0	267
香蕉束顶病 Banana bunchy top	0	1	0	73	0	0	0	0	74
香蕉黑星病 Macrophoma musae	0	0	7	0	134	1	0	0	142
香蕉炭疽病 Banana anthracnose	0	0	2	0	0	59	2	0	63
香蕉黑疫病 Banana black blight	0	0	0	0	2	0	49	0	51
香蕉鞘腐病 Banana sheath rot	0	1	0	0	0	0	0	37	38
总计Total	210	347	272	75	139	61	51	39	1194

下载: 导出CSV

表 4 模型测试精度

Table 4 Model testing accuracy

类别 Category	生产者精度/% Producer’s accuracy	用户精度/% User’s accuracy
健康 Healthy	100	99.5
香蕉枯萎病 Fusarium wilt	99.4	99.1
香蕉叶斑病 Banana leaf spot	96.7	98.5
香蕉束顶病 Banana bunchy top	97.3	98.6
香蕉黑星病 Macrophoma musae	96.4	94.5
香蕉炭疽病 Banana anthracnose	96.7	93.7
香蕉黑疫病 Banana black blight	96.1	96.1
香蕉鞘腐病 Banana sheath rot	94.9	97.4
总体分类精度/% Overall accuracy	98.0

下载: 导出CSV

参考文献(24)

[1]	MARTINELLI F, SCALENGHE R, DAVINO S, et al. Advanced methods of plant disease detection: A review[J]. Agron Sustain Dev, 2015, 35(1): 1-25. doi: 10.1007/s13593-014-0246-1
[2]	黄文江, 师越, 董莹莹, 等. 作物病虫害遥感监测研究进展与展望[J]. 智慧农业, 2019, 1(4): 1-11.
[3]	ALVAREZ A M. Integrated approaches for detection of plant pathogenic bacteria and diagnosis of bacterial diseases[J]. Annu Rev Phytopathol, 2004, 42(8): 339-366.
[4]	李翠玲, 姜凯, 马伟, 等. 基于高光谱的番茄叶片斑潜蝇虫害检测[J]. 光谱学与光谱分析, 2018, 38(1): 253-257.
[5]	POLDER G, BLOK P M, VILLIERS H A C, et al. Potato virus Y detection in seed potatoes using deep learning on hyperspectral images[J]. Front Plant Sci, 2019, 10(1): 1-13.
[6]	杨晨, 董丽芳, 赵海士, 等. 基于模糊判别成分分析法的高光谱作物信息提取与分类[J]. 农业工程学报, 2019, 35(21): 158-165.
[7]	AKHTAR A, KHANUM A, KHAN S A, et al. Automated plant disease analysis (APDA): Performance comparison of machine learning techniques[C]//IEEE. International Conference on Frontiers of Information Technology. New York: IEEE, 2013: 60-65.
[8]	SANNAKKI S S, RAJPUROHIT V S, NARGUND V B, et al. Diagnosis and classification of grape leaf diseases using neural networks[C]//IEEE. Fourth International Conference on Computing, Communications and Networking Technologies. New York: IEEE, 2014: 13-18.
[9]	MOKHTAR U, EL-BENDARY N, HASSANEIN A E, et al. SVM-based detection of tomato leaves diseases[C]// Springer, Cham. Intelligent Systems 2014. Berlin: Springer-Verlag, 2015: 641-652.
[10]	MIZUNO K, TERACHI Y. Architectural study of HOG feature extraction processor for real-time object detection[C]//IEEE. IEEE Workshop on Signal Processing, New York: IEEE, 2012: 197-202.
[11]	KOBAYASHI T. BoF meets HOG: Feature extraction based on histograms of oriented p. d. f. gradients for image classification[C]//IEEE. IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2013: 747-754.
[12]	PUN C M, LEE M C. Extraction of shift invariant wavelet features for classification of images with different sizes[J]. IEEE T Pattern Anal, 2004, 26(9): 1228-1233. doi: 10.1109/TPAMI.2004.67
[13]	ANDREAS K, PRENAFETA-BOLDÚ F X. Deep learning in agriculture: A survey[J]. Comput Electron Agr, 2018, 147(1): 70-90.
[14]	ALVARO F, SOOK Y, KIM S C, et al. A robust deep-learning-based detector for real-time tomato plant diseases and pests recognition[J]. Sensors, 2017, 17(9): 2022-2043. doi: 10.3390/s17092022
[15]	ZHANG X H, QIAO Y, MENG F F, et al. Identification of maize leaf diseases using improved deep convolutional neural networks[J]. IEEE Access, 2018, 6: 370-377.
[16]	陈桂芬, 赵姗, 曹丽英, 等. 基于迁移学习与卷积神经网络的玉米植株病害识别[J]. 智慧农业, 2019, 1(2): 34-44.
[17]	RONNEL R, PARK D. A multiclass deep convolutional neural network classifier for detection of common rice plant anomalies[J]. J Netw Comput Appl, 2018, 9: 67-70.
[18]	RAMCHARAN A, BARANOWSKI K, MCCLOSKEY P, et al. Deep learning for image-based cassava disease detection[J]. Front Plant Sci, 2017, 8: 1852-1859. doi: 10.3389/fpls.2017.01852
[19]	MWEBAZE E, OWOMUGISHA G. Machine learning for plant disease incidence and severity measurements from leaf images[C]//IEEE. International Conference on Machine Learning and Applications. New York: IEEE, 2017: 158-163.
[20]	FERENTINOS K P. Deep learning models for plant disease detection and diagnosis[J]. Comput Electron Agr, 2018, 145: 311-318. doi: 10.1016/j.compag.2018.01.009
[21]	吴华瑞. 基于深度残差网络的番茄叶片病害识别方法[J]. 智慧农业, 2019, 1(4): 42-49.
[22]	SRDJAN S, MARKO A, ANDERLA A, et al. Deep neural networks based recognition of plant diseases by leaf image classification[J]. Comput Intel Neurosc, 2016, 2016: 1-11.
[23]	SI M M, DENG M H, HAN Y. Using deep learning for soybean pest and disease classification in farmland[J]. Journal of Northeast Agricultural University (English Edition), 2019, 26(1): 64-72.
[24]	SZEGEDY C, LIU W, JIA Y Q, et al. Going deeper with convolutions[C]//IEEE. Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2015: 1-9.

施引文献(31)

期刊类型引用(18)

1.	崔紫宁，陈建平，梁丽梅. “交叉融合”：“微生物天然产物化学”的跨界教育模式. 工业微生物. 2024(01): 194-196 . 百度学术
2.	邓杰，尚楠. 芽孢杆菌群体感应系统研究进展. 生物加工过程. 2024(05): 492-499 . 百度学术
3.	乔真，李佳霖，秦松. C6-HSL信号及群体淬灭对海洋聚球藻(Synechococcus)菌藻共栖体系的调控作用. 海洋科学. 2024(09): 52-62 . 百度学术
4.	李凤兰，吴天祥，邓代霞，袁丹丹，江守发. 太子参乙醇提取物对灰树花菌体生长及胞外多糖的影响机理初步研究. 食品与发酵科技. 2023(02): 28-34 . 百度学术
5.	欧凯玉，逄建龙，张一敏，董鹏程，罗欣，毛衍伟. 天然酚类化合物的抑菌作用及在肉与肉制品中的应用研究进展. 食品科学. 2023(09): 358-366 . 百度学术
6.	高鑫，李博. 水产腐败群体感应系统与天然抑菌剂的研究进展. 保鲜与加工. 2023(06): 73-80 . 百度学术
7.	郑爱娟，陈星，张广民，王泽栋，陈志敏，常文环，蔡辉益，刘国华. N-酰基高丝氨酸内酯酶对肉仔鸡生长性能、屠宰性能和养分表观代谢率的影响. 动物营养学报. 2023(06): 3607-3616 . 百度学术
8.	熊儒恒，阎俊，谢晶. 生物被膜初始黏附调控机制及其在食品品质控制中的应用研究进展. 食品科学. 2023(13): 203-215 . 百度学术
9.	杨约萍，高倩倩，宁静，宫佳，胡媛媛，施祖荣. 细菌群体通讯信号及其淬灭研究进展. 仲恺农业工程学院学报. 2022(01): 65-70 . 百度学术
10.	廖才江，李会，王士源，熊静，梅翠，刘丹，何玉张，彭练慈，宋振辉，陈红伟. 生物被膜：益生菌肠道定植的新策略. 生物工程学报. 2022(08): 2821-2839 . 百度学术
11.	乔真，李佳霖，秦松 . 海洋藻际环境中细菌群体感应研究进展. 生物学杂志. 2022(05): 93-97+107 . 百度学术
12.	王亚军，司运美，李彦娟. 群体感应在生物强化功能菌定殖及降解能力增强中的作用研究进展. 应用生态学报. 2022(10): 2871-2880 . 百度学术
13.	李艳群，陈柔雯，林宗豪，田新朋，尹浩. 一株群体感应抑制活性海洋放线菌的筛选与鉴定. 热带海洋学报. 2021(01): 75-81 . 百度学术
14.	杨艳北，许晶，沈城辉，许继国，饶友生. N-酰基高丝氨酸内酯酶的生物信息学分析. 甘肃农业科技. 2021(02): 31-37 . 百度学术
15.	杨艳北，许晶，李袁飞，贡继尚，饶友生. 沼泽红假单胞菌LuxR家族调控蛋白的生物信息学分析. 江苏农业科学. 2021(06): 40-45 . 百度学术
16.	郑钰婷，胡宇如，胡方平，蔡学清. 利用aiiA基因筛选抗烟草青枯病生防菌株及其鉴定. 核农学报. 2021(06): 1322-1328 . 百度学术
17.	宋凯，周莲，何亚文. DSF-家族群体感应信号生物合成途径与调控机制研究进展. 微生物学通报. 2021(04): 1239-1248 . 百度学术
18.	赵祯，肖翎，戚建华，刘韵怡，王年，郁小娟，宋增福. 群体感应淬灭酶YtnP对草鱼肠道菌群结构的影响. 南方农业学报. 2020(11): 2817-2826 . 百度学术