Wheat Disease Recognition Model Integrating Attention Mechanism and Multi-scale Convolution

doi:10.11924/j.issn.1000-6850.casb2023-0216

Abstract

Abstract:

Classification and recognition techniques based on deep convolutional neural networks are proposed to solve the wheat disease identification problem. In this paper, the proposed method is compared and analyzed against existing wheat disease recognition models, and the promising results of applying deep learning technology in agricultural disease identification and monitoring are discussed. The experimental results show that the method based on the deep convolutional neural network is 3%, 4%, 5%, and 5% higher than the traditional EfficientNet network in terms of accuracy, precision, recall, and F₁, and can effectively identify wheat disease. We verified the effectiveness of the proposed components by ablation experiments, and the convergence speed of the model was significantly better than the existing models. The proposed identification model provides strong support for agricultural production and is expected to become one of the main technical tools in the field of agricultural disease identification and monitoring in the future.

Key words: image recognition, disease recognition, deep convolutional neural network, attention mechanism, crop protection

LU Xiaowei, MENG Zhiqing. Wheat Disease Recognition Model Integrating Attention Mechanism and Multi-scale Convolution[J]. Chinese Agricultural Science Bulletin, 2023, 39(25): 147-154.

Figures/Tables 9

References 27

[1]	王东方, 汪军. 基于迁移学习和残差网络的农作物病害分类[J]. 农业工程学报, 2021, 37(4):199-207.
[2]	PFRIEME A K, RUCKWIED B, HABEKUß A, et al. Identification and validation of quantitative trait loci for wheat dwarf virus resistance in wheat (Triticum spp.)[J]. Frontiers in plant science, 2022, 13:483.
[3]	中华人民共和国国家统计局. 国家数据[EB/OL].http://www.stats.gov.cn/tjsj/.
[4]	GAUTAM A K, KUMAR S. Techniques for the detection, identification, and diagnosis of agricultural pathogens and diseases[J]. Natural remedies for pest, disease and weed control, 2020:135-142.
[5]	孙瑜, 张永梅, 武玉军. 基于粒子群算法和支持向量机的黄花菜叶部病害识别[J]. 中国农学通报, 2022, 38(8):135-140. doi: 10.11924/j.issn.1000-6850.casb2021-1215
[6]	DONG S, WANG P, ABBAS K. A survey on deep learning and its applications[J]. Computer science review, 2021, 40:100379. doi: 10.1016/j.cosrev.2021.100379 URL
[7]	TU H, WANG W, CHEN J, et al. Unpaired image-to-image translation with improved two-dimensional feature[J]. Multimedia tools and applications, 2022:1-22.
[8]	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.
[9]	BANERJEE N, BORAH S, SETHI N. Intelligent stuttering speech recognition: A succinct review[J]. Multimedia tools and applications, 2022, 81(17):24145-24166. doi: 10.1007/s11042-022-12817-z
[10]	屠杭垚, 王万良, 陈嘉诚, 等. 结合大气散射模型的生成对抗网络去雾算法[J]. 浙江大学学报(工学版), 2022, 56(2):225-235.
[11]	LV M, ZHOU G, HE M, et al. Maize leaf disease identification based on feature enhancement and DMS-robust alexnet[J]. IEEE Access, 2020, 8:57952-57966. doi: 10.1109/Access.6287639 URL
[12]	LI X, RAI L. Apple leaf disease identification and classification using resnet models[A].// 2020 IEEE 3rd International Conference on Electronic Information and Communication Technology (ICEICT)[C]. IEEE, 2020:738-742.
[13]	LI Miao, WANG Jingxian, LI Hualong, et al. Method for identifying crop disease based on CNN and transfer learning[J]. Smart agriculture, 2019, 1(3):46. doi: 10.12133/j.smartag.2019.1.3.201903-SA005
[14]	LIU L, DONG Y, HUANG W, et al. A disease index for efficiently detecting wheat fusarium head blight using sentinel-2 multispectral imagery[J]. IEEE Access, 2020, 8:52181-52191. doi: 10.1109/Access.6287639 URL
[15]	BHAWARKAR Y, BHURE K, CHAUDHARY V, et al.Diabetic retinopathy detection from fundus images using multi-tasking model with efficient net B5[A].// International Conference on Automation, Computing and Communication 2022[C]. 2022, 44:03027.
[16]	DUAN Y, SUN L, WANG Y. Se-densenet: Attention-based network for detecting pathological images of metastatic breast cancer[A].// Proceedings of the 2019 8th International Conference on Computing and Pattern Recognition[C]. 2019:240-245.
[17]	陈芳芳, 宋姿睿, 张景涵, 等. 融合多尺度特征的冬小麦空间分布提取方法[J]. 农业工程学报, 2022, 38(24):268-274.
[18]	YUAN Y, CHEN L, WU H, et al. Advanced agricultural disease image recognition technologies: A review[J]. Information processing in agriculture, 2022, 9(1):48-59. doi: 10.1016/j.inpa.2021.01.003 URL
[19]	JIANG Y, CHEN L, ZHANG H, et al. Breast cancer histopathological image classification using convolutional neural networks with small SE-ResNet module[J]. PloS one, 2019, 14(3):e0214587.
[20]	亢洁, 刘港, 郭国法. 基于多尺度融合模块和特征增强的杂草检测方法[J]. 农业机械学报, 2022, 53(4):254-260.
[21]	CHEN S, LIU Y, GAO X, et al. Mobilefacenets: Efficient CNNS for accurate real-time face verification on mobile devices[A].// Biometric Recognition: 13th Chinese Conference, CCBR 2018[C]. Springer International Publishing, 2018:428-438.
[22]	连榕榕. 小麦病害分类数据集[DB/OL]. (2022-02-09). https://aistudio.baidu.com/aistudio/datasetdetail/127559.
[23]	葛轶洲, 刘恒, 王言, 等. 小样本困境下的深度学习图像识别综述[J]. 软件学报, 2022, 33(1):18.
[24]	谢星星, 程塨, 姚艳清, 等. 动态特征融合的遥感图像目标检测[J]. 计算机学报, 2022, 45(4):735-747.
[25]	FAN Z, LIN H, LI C, et al. Use of parallel ResNet for high-performance pavement crack detection and measurement[J]. Sustainability, 2022, 14(3):1825. doi: 10.3390/su14031825 URL
[26]	BI C, WANG J, DUAN Y, et al. MobileNet based apple leaf diseases identification[J]. Mobile networks and applications, 2022:1-9.
[27]	SELVARAJU R R, COGSWELL M, DAS A, et al. Grad-CAM: Visual explanations from deep networks via gradient-based localization[A]. //Proceedings of the IEEE international conference on computer vision[C]. 2017:618-626.

类别	数量	英文名称
正常小麦	652	Healthy Wheat
叶锈病	833	Leaf Rust
白粉病	348	Powdery Mildew
小麦黑穗病	832	Wheat Smut
小麦根腐病	468	Wheat Root Rot
小麦赤霉病	484	Wheat Scab
小麦叶枯病	470	Wheat Leaf Blotch

类别	数量	英文名称
正常小麦	652	Healthy Wheat
叶锈病	833	Leaf Rust
白粉病	348	Powdery Mildew
小麦黑穗病	832	Wheat Smut
小麦根腐病	468	Wheat Root Rot
小麦赤霉病	484	Wheat Scab
小麦叶枯病	470	Wheat Leaf Blotch

模型	准确率	精确率	召回率	F₁
MobileNet	0.84	0.81	0.80	0.80
ResNet	0.87	0.85	0.84	0.85
EfficientNet	0.89	0.87	0.86	0.86
EfficientNet-SE-MSC	0.92	0.91	0.91	0.91

模型	准确率	精确率	召回率	F₁
MobileNet	0.84	0.81	0.80	0.80
ResNet	0.87	0.85	0.84	0.85
EfficientNet	0.89	0.87	0.86	0.86
EfficientNet-SE-MSC	0.92	0.91	0.91	0.91

模型	准确率	精确率	召回率	F₁
EfficientNet	0.89	0.87	0.86	0.86
With MSC	0.91	0.89	0.86	0.88
With SE	0.90	0.88	0.89	0.88
EfficientNet-SE-MSC	0.92	0.91	0.91	0.91