张耀光,
刘涛,
河北科技大学经济管理学院 石家庄 050018
基金项目: 河北省重点研发计划项目19226417D
河北省高等学校科学技术重点项目ZD2019083
详细信息
作者简介:杨明欣, 主要从事信息管理、信息安全方面的研究。E-mail:ymxspj@163.com
通讯作者:刘涛, 主要研究方向为信息资源管理和大数据分析建模。E-mail:liutaolunwen@163.com
中图分类号:TP183计量
文章访问数:191
HTML全文浏览量:12
PDF下载量:106
被引次数:0
出版历程
收稿日期:2020-05-20
录用日期:2020-09-26
刊出日期:2020-12-01
Corn disease recognition based on the Convolutional Neural Network with a small sampling size
YANG Mingxin,ZHANG Yaoguang,
LIU Tao,
School of Economics and Management, Hebei University of Science and Technology, Shijiazhuang 050018, China
Funds: the Key R & D Program of Hebei Province of China19226417D
the Key Science and Technology Project of Higher School of Hebei Province of ChinaZD2019083
More Information
Corresponding author:LIU Tao, E-mail:liutaolunwen@163.com
摘要
HTML全文
图
参考文献
相关文章
施引文献
资源附件
访问统计
摘要
摘要:农作物病害治理对于农作物的产量和品质有着非常重要的影响。本文针对玉米病害人工识别困难、识别过程耗费大量的人力成本和病害数据样本小且分布不均的问题，提出了一种改进的迁移学习神经网络（Neural Network）的病害识别方法。首先，采用旋转、翻转等方法对样本图像集进行数据增强；其次，通过迁移的MobileNetV2模型在玉米病害图像数据集上训练，利用Focal Loss函数改进神经网络的损失函数；最后，通过Softmax分类方法实现玉米病害图像识别。另外通过试验对比AlexNet、GooleNet、Vgg16、RestNet34、MobileNetV2和迁移的MobileNetV2这6种模型的训练集准确率、验证集准确率、权重、参数数量和运行时间。结果显示，6种模型验证集的准确率分别为93.88%、95.48%、91.69%、97.67%、96.21%和97.23%，迁移的MobileNetV2的准确率最高，且权重仅有8.69 MB。进一步通过混淆矩阵对比了MobileNetV2和迁移的MobileNetV2两种模型，迁移的MobileNetV2模型识别正确率提升1.02%，训练速度减少6 350 s。本文提出迁移的MobileNetV2模型对玉米病害小样本的识别效果最佳，具备更好的收敛速度与识别能力，同时能够降低模型的运算量并大幅度缩短识别时间。
关键词:玉米病害/
迁移学习/
小样本/
卷积神经网络/
Focal Loss/
混淆矩阵
Abstract:Crop disease management influences yield and quality, yet identifying corn diseases is still difficult. High labor costs, small number of sample, and uneven disease distributions contribute to the difficulty. We propose an improved Convolutional Neural Network (CNN) model based on the transfer learning method for disease identification. The sample image set was enhanced by rotation and roll-over, then the migrated MobileNetV2 model was used to train the image data set for corn diseases. The Focal Loss function was used to improve the neural network loss function, and the Softmax classification method was used for corn disease image recognition. The training set accuracy, validation set accuracy, weight, run time, and the number of parameter in six models were experimentally compared. The verification set accuracy rates were 93.88% (AlexNet), 95.48% (GoogleNet), 91.69% (Vgg16), 97.67% (RestNet34), 96.21% (MobileNetV2), and 97.23% (migrated MobileNetV2). The migrated MobileNetV2 was 97.23% accurate and weighed 8.69 MB. Confounding the MobileNetV2 model improved the recognition accuracy by 1.02% and reduced the training speed by 6 350 seconds compared to the unconfounded model. The migrated MobileNetV2 model had the best corn disease recognition ability with a small sampling size; improved convergence speed, reduced model calculations, and greatly improved the recognition time.
Key words:Corn diseases and insect pests/
Transfer learning/
Small sample/
Convolutional Neural Network/
Focal Loss/
Confusion matrix
HTML全文
图1残差网络(a)和倒残差网络(b)结构图
Figure1.Structures of the residual network (a) and the inverted residual network (b)
下载: 全尺寸图片幻灯片
图2第1次卷积后形成的32个子图(a)和第1次残差网络后形成的16个子图(b)
Figure2.Thirty-two subgraphs formed after the first convolution (a) and 16 subgraphs formed after the first residual network (b)
下载: 全尺寸图片幻灯片
图3预处理后5种玉米病害类型的效果图(从左向右依次为玉米矮花叶病毒病、玉米灰斑病、玉米锈病、健康玉米、玉米叶斑病)
Figure3.Pre-processed pictures of maize (from left to right) dwarf mosaic disease, maize gray leaf spots, maize rust, healthy maize and maize leaf spots
下载: 全尺寸图片幻灯片
图4MobileNetV2训练集(a)和迁移的MobileNetV2训练集(b)的混淆矩阵
标签0、1、2、3、4和5分别对应健康玉米、玉米灰斑病、玉米锈病、玉米叶斑病、玉米矮花叶病毒病。
Figure4.Confusion matrixes of MobileNetV2 training set (a) and migrated MobileNetV2 training set (b)
Label 0, 1, 2, 3, 4 and 5 correspond to healthy maize, maize gray leaf spots, maize rust, maize leaf spots and maize dwarf mosaic disease, respectively.
下载: 全尺寸图片幻灯片
图5MobileNetV2验证集的混淆矩阵(a)和迁移的MobileNetV2的混淆矩阵(b)
标签0、1、2、3、4和5分别对应健康玉米、玉米灰斑病、玉米锈病、玉米叶斑病、玉米矮花叶病毒病。
Figure5.Confusion matrixes of the MobileNetV2 verification set (a) and migrated MobileNetV2 verification set (b)
Label 0, 1, 2, 3, 4 and 5 correspond to healty maize, maize gray leaf spots, maize rust, maize leaf spots and maize dwarf mosaic disease, respectively.
下载: 全尺寸图片幻灯片
图66种模型训练集进行50次迭代的损失曲线
Figure6.Loss curves of the training set for 50 iterations of six models
下载: 全尺寸图片幻灯片
图76种模型验证集进行50次迭代的损失曲线
Figure7.Loss curves of the validation set for 50 iterations of six models
下载: 全尺寸图片幻灯片表1Dropout方法中概率p选择
Table1.Probability p selection in the Dropout method
| p | 训练集准确率 Training set accuracy (%) | 测试集准确率 Valid set accuracy (%) |
| 0.1 | 92.52 | 95.77 |
| 0.2 | 93.53 | 97.23 |
| 0.3 | 92.93 | 96.06 |
| 0.4 | 93.01 | 95.48 |
| 0.5 | 92.88 | 95.36 |
| 0.6 | 92.33 | 95.77 |
| 0.7 | 93.31 | 95.59 |
| 0.8 | 91.82 | 95.04 |
| 0.9 | 89.64 | 94.90 |
下载: 导出CSV表2数据增强后的玉米病害数据集
Table2.Maize diseases dataset after data enhancement
| 标签 Label | 标签名称 Label name | 训练集数量 Number of training set | 增强后训练集数量 Number of training sets after enhancement |
| 0 | 健康玉米?Corn healthy | 320 | 640 |
| 1 | 玉米灰斑病?Maize gray leaf spots | 358 | 716 |
| 2 | 玉米锈病?Maize rust | 838 | 838 |
| 3 | 玉米叶斑病?Maize leaf spots | 669 | 669 |
| 4 | 玉米矮花叶病毒病?Maize dwarf mosaic virus | 815 | 815 |
| ??标签0-4的意义见表 2。The meaning of the lable 0-4 is shown in the table 2. | |||
下载: 导出CSV表35种玉米病害类型训练集准确率
Table3.Train set accuracies of five diseases of maize
| 训练集 Training set | 标签0准确率 Label 0 accuracy (%) | 标签1准确率 Label 1 accuracy (%) | 标签2准确率 Label 2 accuracy (%) | 标签3准确率 Label 3 accuracy (%) | 标签4准确率 Label 4 accuracy (%) | 总训练集的准确率 Total training set accuracy (%) |
| 原始?Original | 98.42 | 84.56 | 98.06 | 82.72 | 96.89 | 92.77 |
| 增强?Enhancement | 99.38 | 89.19 | 97.62 | 85.71 | 99.39 | 94.62 |
| ??标签0-4的意义见表 2。The meaning of the lable 0-4 is shown in the table 2. | ||||||
下载: 导出CSV表46种模型进行玉米病害识别的测试结果
Table4.Recognition results of maize diseases by six models
| 模型名称 Model name | 训练集准确率 Training set accuracy (%) | 验证集准确率 Validation set accuracy (%) | 权重大小 Weight size (MB) | 参数数量 Number of parameters | 运行时间 Run time (s) |
| AlexNet | 95.24 | 93.88 | 55.67 | 14 591 685 | 2 830 |
| GooleNet | 96.38 | 95.48 | 39.39 | 10 318 655 | 10 400 |
| Vgg16 | 95.05 | 91.69 | 158.17 | 41 460 549 | 51 800 |
| RestNet34 | 96.17 | 97.67 | 81.31 | 21 287 237 | 12 550 |
| MobileNetV2 | 94.86 | 96.21 | 8.69 | 2 230 277 | 9 050 |
| Migrated MobileNetV2 | 94.62 | 97.23 | 8.69 | 2 230 277 | 2 700 |
下载: 导出CSV参考文献
| [1] | 李道亮.智慧农业:中国的机遇和挑战[J].高科技与产业化, 2015, 11(5):42-45 LI D L. Smart agriculture:Opportunities and challenge for China[J]. High-Technology and Industrialization, 2015, 11(5):42-45 |
| [2] | 李冠林, 马占鸿, 王海光.基于支持向量机的小麦条锈病和叶锈病图像识别[J].中国农业大学学报, 2012, 17(2):72-79 LI G L, MA Z H, WANG H G. Image recognition of wheat stripe rust and wheat leaf rust based on support vector machine[J]. Journal of China Agricultural University, 2012, 17(2):72-79 |
| [3] | 李震, 洪添胜, 曾祥业, 等.基于K-means聚类的柑橘红蜘蛛图像目标识别[J].农业工程学报, 2012, 28(23):147-153 LI Z, HONG T S, ZENG X Y, et al. Citrus red mite image target identification based on K-means clustering[J]. Transactions of the Chinese Society of Agricultural Engineering, 2012, 28(23):147-153 |
| [4] | 许良凤, 徐小兵, 胡敏, 等.基于多分类器融合的玉米叶部病害识别[J].农业工程学报, 2015, 31(14):194-201 XU L F, XU X B, HU M, et al. Corn leaf disease identification based on multiple classifiers fusion[J]. Transactions of the Chinese Society of Agricultural Engineering, 2015, 31(14):194-201 |
| [5] | Turko?lu M, Hanbay D. Plant disease and pest detection using deep learning-based features[J]. Turkish Journal of Electrical Engineering and Computer Sciences, 2019, 27(3):1636-1651 doi: 10.3906/elk-1809-181 |
| [6] | Lecun Y, Bottou L. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11):2278-2324 doi: 10.1109/5.726791 |
| [7] | KRIZHEVSKY A, SUTSKEVE R I, HINTON G E. ImageNet classification with deep convolutional neural networks[C]//Proceedings of the 25th International Conference on Neural Information Processing Systems. La Jolla, CA: Neural Information Processing Systems Foundation, 2012: 1097-1105 |
| [8] | SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[EB/OL].[2019-02-10]. https://arxiv.org/pdf/1409.1556.pdf. |
| [9] | SZEGEDY C, LIU W, JIA Y Q, et al. Going deeper with convolutions[C]//2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Boston, MA, USA: IEEE, 2015: 1-9 |
| [10] | HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, NV, USA: IEEE, 2016: 770-778 |
| [11] | MOHANTY S P, HUGHES D P, SALATHé M. Using deep learning for image-based plant disease detection[J]. Frontiers in Plant Science, 2016, 7:1419 doi: 10.3389/fpls.2016.01419 |
| [12] | WANG G, SUN Y, WANG J X. Automatic image-based plant disease severity estimation using deep learning[J]. Computational Intelligence and Neuroscience, 2017, 2017:2917536 |
| [13] | Oppenheim D, Shani G, Erlich O, et al. Using deep learning for image-based potato tuber disease detection[J]. Phytopathology, 2019, 109(6):1083-1087 doi: 10.1094/PHYTO-08-18-0288-R |
| [14] | SLADOJEVIC S, ARSENOVIC M, ANDERLA A, et al. Deep neural networks based recognition of plant diseases by leaf image classification[J]. Computational Intelligence and Neuroscience, 2016, 2016:3289801 |
| [15] | 李勇, 刘战东, 张海军.不平衡数据的集成分类算法综述[J].计算机应用研究, 2014, 31(5):1287-1291 LI Y, LIU Z D, ZHANG H J. Review on ensemble algorithms for imbalanced data classification[J]. Application Research of Computers, 2014, 31(5):1287-1291 |
| [16] | 刘悦婷, 孙伟刚, 张发菊.一种新的近邻密度SVM不平衡数据集分类算法[J].贵州大学学报:自然科学版, 2019, 36(3):75-80 LIU Y T, SUN W G, ZHANG F J. Imbalanced dataset classification algorithm based on NNDSVM[J]. Journal of Guizhou University:Natural Sciences, 2019, 36(3):75-80 |
| [17] | IOFFE S, SZEGEDY C. Batch normalization: Accelerating deep network training by reducing internal covariate shift[C]//Proceedings of the 32nd International Conference on International Conference on Machine Learning (ICML). Lille: ACM, 2015: 448-456 |
| [18] | PAN S J, YANG Q. A survey on transfer learning[J]. IEEE Transactions on Knowledge and Data Engineering, 2010, 22(10):1345-1359 doi: 10.1109/TKDE.2009.191 |
| [19] | HINTON G E, SRIVASTAVA N, KRIZHEVSKY A, et al. Improving neural networks by preventing co-adaptation of feature detectors[J]. Computerence, 2012, 3(4):212-223 |
| [20] | LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//2017 IEEE International Conference on Computer Vision (ICCV). Venice, Italy: IEEE, 2017: 2999-3007 |
