基于改进ResNet34的辣椒病虫害图像识别

Image Recognition of Pepper Diseases and Pests Based on Improved ResNet34

摘要

针对辣椒病虫害图像识别任务中模型计算复杂度高、参数规模大、难以在资源受限设备上部署的问题,提出一种基于改进ResNet34的轻量化病虫害识别方法.首先,通过引入深度可分离卷积结构替换传统标准卷积,在保证特征提取能力的同时显著降低模型参数量和计算开销;其次,融合卷积块注意力机制,从通道和空间两个维度增强模型对病斑关键区域特征的关注能力,提高特征表达的判别性.为验证该方法的有效性,在辣椒病虫害数据集及公开PlantVillage番茄病害数据集上进行系统实验.实验结果表明,改进模型在多项评价指标上均取得显著提升,其中识别准确率达98.36%,较原始ResNet34提高5.64%.在与MobileNetV2,GoogLeNet,AlexNet,VGG16和ResNet18等模型的对比实验中,该方法在准确率、精确率、召回率和F₁值等指标上均表现更优.此外,模型的参数量仅为2.73M,模型体积为10.62MiB,在保持较高识别精度的同时具有良好的轻量化特性和部署适用性.

Abstract

Aiming at the problems of high computational complexity, large parameter size, and limited deployability of existing models in the image recognition tasks of pepper diseases and pests, we proposed a lightweight disease and pest recognition method based on an improved ResNet34. Firstly, depthwise separable convolution structures were introduced to replace conventional standard convolutions, which significantly reduced model parameter count and computational cost while maintaining feature extraction capability. Secondly, a convolutional block attention mechanism was integrated to enhance the model’s ability to focus on critical region features of lesions from both channel and spatial dimensions, and improve the discriminative power of feature representations. In order to verify the effectiveness of the proposed method, systematic experiments were conducted on a pepper disease and pest dataset as well as the public PlantVillage tomato disease dataset. Experimental results show that the improved model achieves significant improvements in multiple evaluation metrics, with a recognition accuracy of 98.36%, which is 5.64% higher than the original ResNet34. In comparative experiments with models such as MobileNetV2, GoogLeNet, AlexNet, VGG16, and ResNet18, the proposed method performs better in accuracy, precision, recall, and F1 values. Furthermore, the parameter count of proposed model is only 2.73M, and the model volume is 10.62 MiB, which has good lightweight characteristics and deployment applicability while maintaining high recognition accuracy.

关键词

辣椒病虫害 / ResNet34网络 / 深度可分离卷积 / 注意力机制 / 深度学习

Key words

pepper disease and pest / ResNet34 network / depthwise separable convolution / attention mechanism / deep learning

参考文献

[1] 李玲, 王红, 刘庆生, 等. 基于纹理特征和支持向量机的ALOS图像土地覆被分类[J]. 国土资源遥感, 2011(4):58-63.

[2] (Li L, Wang H, Liu Q S, et al. Land Cover Classification of ALOS Images Based on Texture Features and Support Vector Machine[J]. Remote Sensing for Land and Resources, 2011(4):58-63.)

[3] 魏丽冉, 岳峻, 李振波, 等. 基于核函数支持向量机的植物叶部病害多分类检测方法[J]. 农业机械学报, 2017, 48(增刊):166-171.

[4] (Wei L R, Yue J, Li Z B, et al. Multi-classification Detection Method for Plant Leaf Diseases Based on Kernel Function Support Vector Machine[J]. Transactions of the Chinese Society for Agricultural Machinery, 2017, 48(Suppl):166-171.)

[5] Qadri S, Qadri S F, Husnain M, et al. Machine Vision Approach for Classification of Citrus Leaves Using Fused Features[J]. International Journal of Food Properties, 2019, 22(1):2072-2089.

[6] Hirasawa T, Aoyama K, Tanimoto T, et al. Application of Artificial Intelligence Using a Convolutional Neural Network for Detecting Gastric Cancer in Endoscopie Images[J]. Gastric Cancer, 2018, 21(4):653-660.

[7] 严律南. 人工智能在医学领域应用的现状与展望[J]. 中国普外基础与临床杂志, 2018, 25(5):513-514.

[8] (Yan L N. Current Status and Prospects of Artificial Intelligence Applications in the Medical Field[J]. Chinese Journal of Bases and Clinics in General Surgery, 2018, 25(5):513-514.)

[9] 王刚, 陈红, 马敬玲, 等. 基于多尺度1D-CNN和注意力机制的汇率多步预测研究[J]. 系统工程理论与实践, 2024, 44(6):1934-1949.

[10] (Wang G, Chen H, Ma J L, et al. Multi-step Exchange Rate Prediction Based on Multi-scale 1D-CNN and Attention Mechanism[J]. Systems Engineering:Theory & Practice, 2024, 44(6):1934-1949.)

[11] Juan C, Rivera P, Christian B, et al. Factors Affecting Deep Learning Model Performance in Citizen Science-Based Image Data Collection for Agriculture:A Case Study on Coffee Crops[J]. Computers and Electronics in Agriculture, 2025, 232:110096-1-110096-13.

[12] Alruwaili M, Alanazi S, El-Ghany A S, et al. An Efficient Deep Learning Model for Olive Diseases Detection[J]. International Journal of Advanced Computer Science and Applications, 2019, 10(8):486-492.

[13] 赵法川, 徐晓辉, 宋涛, 等. 融合多头注意力的轻量级作物病虫害识别[J]. 华南农业大学学报, 2023, 44(6):986-994.

[14] (Zhao F C, Xu X H, Song T, et al. Lightweight Crop Pest and Disease Recognition Integrating Multi-head Attention Mechanism[J]. Journal of South China Agricultural University, 2023, 44(6):986-994.)

[15] 戴敏, 孙文靖, 缪宏. 基于轻量化CBAM-GoogLeNet的辣椒病虫害识别[J]. 中国农机化学报, 2025, 46(2):224-229.

[16] (Dai M, Sun W J, Miao H. Pepper Pest and Disease Recognition Based on Lightweight CBAM-GoogLeNet[J]. Journal of Chinese Agricultural Mechanization, 2025, 46(2):224-229.)

[17] Ferentinos K P. Deep Learning Models for Plant Disease Detection and Diagnosis[J]. Computers and Electronics in Agriculture, 2018, 145:311-318.

[18] 谢琬, 崔艳荣. 基于改进EfficientNetV2模型的玉米叶片病害识别[J]. 江苏农业科学, 2025, 53(9):207-215.

[19] (Xie W, Cui Y R. Corn Leaf Disease Recognition Based on Improved EfficientNetV2 Model[J]. Jiangsu Agricultural Sciences, 2025, 53(9):207-215.)

[20] He K M, Zhang X Y, Ren S Q, et al. Deep Residual Learning for Image Recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE, 2016:770-778.

[21] Howard A G. Mobilenets:Efficient Convolutional Neural Networks for Mobile Vision Applications[EB/OL]. (2017-04-17)[2025-01-10]. https://arxiv.org/abs/1704.04861.

[22] 张缓缓, 冯屹轩, 吴宏超. 基于改进ShuffleNetV2网络的路面类型识别[J]. 江苏大学学报(自然科学版), 2026, 47(1):48-54.

[23] (Zhang H H, Feng Y X, Wu H C. Pavement Type Recognition Based on Improved ShuffleNetV2 Network[J]. Journal of Jiangsu University(Natural Science Edition), 2026, 47(1):48-54.)

[24] 孙伟, 陈俊杰. MF-cache:用于玉米病害识别的CLIP多模态缓存模型[J]. 计算机工程, 2026, 52(3):420-428.

[25] (Sun W, Chen J J. MF-cache:A CLIP Multimodal Cache Model for Maize Disease Identification[J]. Computer Engineering, 2026, 52(3):420-428.)

[26] Woo S, Park J, Lee J Y, et al. CBAM:Convolutional Block Attention Module[C]// Proceedings of the European Conference on Computer Vision.Berlin:Springer, 2018:3-19.

[27] 王美华, 吴振鑫, 周祖光. 基于注意力改进CBAM的农作物病虫害细粒度识别研究[J]. 农业机械学报, 2021, 52(4):239-247.

[28] (Wang M H, Wu Z X, Zhou Z G. Fine-Grained Crop Pest and Disease Recognition Using Attention-Enhanced CBAM[J]. Transactions of the Chinese Society for Agricultural Machinery, 2021, 52(4):239-247.)

[29] Wang Z Q, Zhou W X, Li Y M. GFN:A Garbage Classification Fusion Network Incorporating Multiple Attention Mechanisms[J]. Electronics, 2025, 14(1):75-1-75-22.

[30] 易鹏, 杨晔, 严仕嘉. 基于MPCNN模型的sEMG快速迁移学习的手势识别应用研究[J]. 计算机工程, 2025, 51(1):304-311.

[31] (Yi P, Yang Y, Yan S J. Research on Gesture Recognition Application Based on MPCNN Model for Fast Transfer Learning of sEMG[J]. Computer Engineering, 2025, 51(1):304-311.)

[32] Sandler M, Howard A, Zhu M L, et al. MobileNetV2:Inverted Residuals and Linear Bottlenecks[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE, 2018:4510-4520.

[33] Szegedy C, Liu W, Jia Y, et al. Going Deeper with Convolutions[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Berlin:Springer, 2015:1-9.

[34] Perumal A, Nithiyanantham J, Nagaraj J. An Improved AlexNet Deep Learning Method for Limb Tumor Cancer Prediction and Detection[J]. Biomedical Physics & Engineering Express, 2024, 11(1):015004-1-015004-15.

[35] Simonyan K, Zisserman A. Very Deep Convolutional Networks for Large-Scale Image Recognition[EB/OL]. (2014-09-04)[2024-12-30]. https://arxiv.org/abs/1409.1556.

[36] Selvaraju R R, Cogswell M, Das A, et al. Grad-CAM:Visual explanations from Deep Networks via Gradient-Based Localization[J]. International Journal of Computer Vision, 2020, 128(2):336-359.

基金资助

甘肃省基础研究创新群体项目(25JRRA805)