[1]刘翱宇,吴云志,朱小宁,等.基于深度残差网络的玉米病害识别[J].江苏农业学报,2021,(01):67-74.[doi:doi:10.3969/j.issn.1000-4440.2021.01.009]
　LIU Ao-yu,WU Yun-zhi,ZHU Xiao-ning,et al.Corn disease recognition based on deep residual network[J].,2021,(01):67-74.[doi:doi:10.3969/j.issn.1000-4440.2021.01.009]
点击复制

基于深度残差网络的玉米病害识别()
分享到：

江苏农业学报[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2021年01期
页码:
67-74
栏目:
植物保护
出版日期:
2021-02-28

文章信息/Info

Title:
Corn disease recognition based on deep residual network
作者:
刘翱宇1; 吴云志1; 2; 朱小宁3; 范国华1; 2; 乐毅1; 2; 张友华1; 2
（1.安徽农业大学信息与计算机学院，安徽合肥230036；2.安徽省北斗精准农业信息工程实验室，安徽合肥230036；3.安徽农业大学茶与食品科技学院，安徽合肥230036）
Author(s):
LIU Ao-yu1; WU Yun-zhi1; 2; ZHU Xiao-ning3; FAN Guo-hua1; 2;  YUE Yi1; 2; ZHANG You-hua1; 2
(1.School of Information and Computer Science, Anhui Agricultural University, Hefei 230036, China;2.Anhui Provincial Engineering Laboratory for Beidou Precision Agriculture Information, Hefei 230036, China;3.School of Tea and Food Technology, Anhui Agricultural University, Hefei 230036, China)
关键词:
病害图像识别; 深度残差网络; 迁移学习
Keywords:
disease image recognition; deep residual network; transfer learning
分类号:
TP391
DOI:
doi:10.3969/j.issn.1000-4440.2021.01.009
文献标志码:
A
摘要:
针对人工诊断玉米病害成本高、效率低、时延长等问题，提出一种基于深度残差网络的玉米病害识别网络TFL-ResNet。TFL-ResNet网络基于ResNet50网络，首先引入Focal Loss损失函数使模型专注于难分类的病害样本，其次将ResNet50网络在PlantVillage数据集训练好的参数迁移到改进网络上以完成构建。采用的玉米病害数据集涉及健康植株、大斑病、灰斑病、锈病4种标签，并使用旋转、翻转、平移等操作对数据集进行数据增强与扩充。对数据集进行训练和测试，与VGG16等对照模型相比，TFL-ResNet网络收敛速度更快、分类效果更好，平均识别准确率高达98.96%。通过观察精准率、召回率、混淆矩阵等评价指标得出TFL-ResNet网络具有较好的鲁棒性和泛化能力，可用于玉米病害智能诊断。
Abstract:
Aiming at the problems of high cost, low efficiency, time extension and so on for manual diagnosis of corn diseases, a corn disease identification network TFL-ResNet based on deep residual network was proposed. The TFL-ResNet was based on the ResNet50. Firstly, the focal loss function was introduced to make the model focus on disease samples which were difficult to classify. Secondly, the parameters of ResNet50 trained in PlantVillage data set were migrated to the improved network to complete the construction. The corn disease data set used in this study involved four labels: healthy plants, leaf blight, gray leaf spot and rust. Rotation, flip, translation and other operations were used to enhance and expand the data set. Compared with VGG16 and other control models, TFL-ResNet had faster convergence speed and better classification effect. The average recognition accuracy of TFL-ResNet was 98.96%. By observing the model evaluation indicators such as precision rate, recall rate, confusion matrix, TFL-ResNet has good robustness and generalization ability, which can be used for intelligent diagnosis of corn diseases.

参考文献/References:

[1]孟颖,陈桂芬,卢建,等. Simulink平台在玉米病害视频图像中的实时诊断[J]. 吉林农业大学学报, 2017, 39(4): 483-487.
[2]许景辉,邵明烨,王一琛,等. 基于迁移学习的卷积神经网络玉米病害图像识别[J]. 农业机械学报, 2020, 51(2): 230-236，253.
[3]龙满生,欧阳春娟,刘欢,等. 基于卷积神经网络与迁移学习的油茶病害图像识别[J]. 农业工程学报, 2018, 34(18): 194-201.
[4]周云成,许童羽,邓寒冰,等. 基于自监督学习的番茄植株图像深度估计方法[J]. 农业工程学报, 2019, 35(24): 173-182.
[5]孙俊,何小飞,谭文军,等. 空洞卷积结合全局池化的卷积神经网络识别作物幼苗与杂草[J]. 农业工程学报, 2018, 34(11): 159-165.
[6]刘永波,雷波,曹艳,等. 基于深度卷积神经网络的玉米病害识别[J]. 中国农学通报, 2018, 34(36): 159-164.
[7]顾博,邓蕾蕾,李巍,等. 基于GrabCut算法的玉米病害图像识别方法研究[J]. 中国农机化学报, 2019, 40(11): 143-149.
[8]张善文,张传雷. 基于局部判别映射算法的玉米病害识别方法[J]. 农业工程学报, 2014, 30(11): 167-172.
[9]HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]//IEEE. 2016 IEEE conference on computer vision and pattern recognition. Las Vegas: IEEE, 2016: 770-778.
[10]JUNG H, CHOI M K, JUNG J, et al. ResNetbas-ed vehicle classification and localization in traffic surveillance systems[C]//IEEE. 2017 IEEE conference on computer vision and pattern recognition. Honolulu: IEEE, 2017: 61-67.
[11]LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//IEEE. 2017 IEEE conference on computer vision and pattern recognition. Honolulu: IEEE, 2017: 2980-2988.
[12]MOHANTY S P, HUGHES D P, SALATHE M. Using deep learning for image-based plant disease detection[J]. Frontiers in Plant Science, 2016, 7: 1419.
[13]SINGKEK SCI-TECH. Agricultural Disease. 2018 AI Challenger [DB/OL]. (2018-12-19)
[2020-3-20]. https://challenger.ai/competition/pdr2018.
[14]朱威,屈景怡,吴仁彪. 结合批归一化的直通卷积神经网络图像分类算法[J]. 计算机辅助设计与图形学学报, 2017, 29(9): 1650-1657.
[15]PAN S J, YANG Q. A survey on transfer learning[J]. IEEE Transactions on Knowledge and Data Engineering, 2009, 22(10): 1345-1359.
[16]石祥滨,房雪键,张德园,等. 基于深度学习混合模型迁移学习的图像分类[J]. 系统仿真学报, 2016, 28(1): 167-173,182.
[17]任胜男,孙钰,张海燕,等. 基于one-shot学习的小样本植物病害识别[J]. 江苏农业学报, 2019, 35(5): 1061-1067.
[18]HU K, ZHANG Z, NIU X, et al. Retinal vessel segmentation of color fundus images using multiscale convolutional neural network with an improved cross-entropy loss function[J]. Neurocomputing, 2018, 309: 179-191.
[19]SANDLER M, HOWARD A, ZHU M, et al. Mobilenetv2: Inverted residuals and linear bottlenecks[C]//IEEE. 2018 IEEE conference on computer vision and pattern recognition. Salt Lake City: IEEE, 2018: 4510-4520.
[20]SZEGEDY C, VANHOUCKE V, IOFFE S, et al.Rethinking the inception architecture for computervision[C]//IEEE. 2016 IEEE conference on computer vision and pattern recognition. Las Vegas: IEEE, 2016: 2818-2826.
[21]CHOLLET F. Xception: Deep learning with dept-hwise separable convolutions[C]//IEEE. 2017 IEEE Conference on computer vision and pattern recognition. Honolulu: IEEE, 2017: 1251-1258.
[22]刘小峰,舒仁杰,柏林,等. 一种新的稀疏分类融合方法及其在机车轴承故障诊断中的应用[J]. 中国电机工程学报, 2020, 40(17): 5675-5682.

备注/Memo

备注/Memo:
收稿日期：2020-04-15基金项目：国家重点研发计划项目（2017YFD0301303）；安徽省科技重大专项（18030901029）；安徽省高校自然科学研究项目（KJ2019A0211）；安徽省大学生创新创业训练计划项目（S202010364210）作者简介：刘翱宇(1999-)，男，安徽亳州人，本科，研究方向为计算机视觉、图像识别。(E-mail)liuaoyu@ahau.edu.cn。吴云志为共同第一作者。通讯作者：张友华，(E-mail)zhangyh@ahau.edu.cn
更新日期/Last Update: 2021-03-15