融合沙漏结构与改进坐标注意力的轻量级番茄叶片病害识别模型  

A lightweight tomato leaf disease recognition model integrating a sandglass structure with improved coordinate attention

作　　者：谷瑞 宋翠玲[1] 钱春花 GU Rui;SONG Cuiling;QIAN Chunhua(School of Digital Economy and Management,Nanjing University,Nanjing 210008,China;College of Financial Technology,Suzhou Industrial Park Institute of Services Outsourcing,Suzhou 215123,Jiangsu,China;College of Smart Agriculture,Suzhou Polytechnic Institute of Agriculture,Suzhou 215008,China)

机构地区：[1]南京大学数字经济与管理学院,江苏南京210008 [2]苏州工业园区服务外包职业学院金融科技学院,江苏苏州215123 [3]苏州农业职业技术学院智慧农业学院,江苏苏州215008

出　　处：《浙江农业学报》2025年第1期217-230,共14页Acta Agriculturae Zhejiangensis

基　　金：2023年江苏省高职院校教师专业带头人高端研修项目(2023TDFX010);苏州市科技计划(SNG2023005);江苏现代农业产业技术体系项目(JATS-2023-348)。

摘　　要：针对现有番茄叶片识别模型参数量大、计算复杂度高,推理时间长,难以部署在资源受限的移动设备上的问题,本文提出一种轻量级识别网络SG-ICA-MobileNetV3。首先,引入沙漏结构对MobileNetV3Small的倒残差块进行改造,在高维空间建立特征转换和跳跃连接缓解信息丢失问题,强化模型特征学习能力;其次,嵌入改进的坐标注意力机制,融合全局平均池化和最大池化自适应地学习不同位置的特征权重,增强对病害区域的感知能力;最后,将ReLU激活函数替换为ELU,缓解模型训练中梯度消失和权重偏置更新失效现象,提升网络收敛速度。结果表明,该模型在测试集上的分类准确率高达98.36%,在参数量、计算复杂度、推理速度、识别精度等方面优于MobileNetV3Small、MobileNeXt-1.0、MobileVit-S、ConvNeXt-V2等轻量级模型,并具有较强的泛化能力,能为快速、准确识别植物叶片病害提供算法支持。In response to the problems of large parameter sizes,high computational complexity,long inference time,and difficulty in deployment on resource-limited mobile devices associated with existing tomato leaf recognition models,this paper proposes a lightweight recognition network called SG-ICA-MobileNetV3.Firstly,a sandglass structure was introduced to modify the inverted residual blocks of MobileNetV3Small,establishing feature transformations and skip connections in high-dimensional space to mitigate information loss and strengthen the model’s feature learning capability.Secondly,an improved coordinate attention mechanism was embedded,integrating global average pooling and max pooling to adaptively learn feature weights at different positions,enhancing the perception ability of diseased areas.Finally,the ReLU activation function was replaced with ELU to alleviate gradient vanishing and the issue of weight bias updated failure during model training,improving network convergence speed.The results showed that the model achieved a classification accuracy of 98.36%on the test set.It outperformed lightweight models such as MobileNetV3Small,MobileNeXt-1.0,MobileVit-S,and ConvNeXt-V2 in terms of parameter count,computational complexity,inference speed,and recognition accuracy,demonstrating strong generalization capabilities.This study result could provide algorithm support for fast and accurate identification of plant leaf diseases.

关 键 词：沙漏结构 SG-ICA-MobileNetV3 坐标注意力 番茄 叶片 病害识别 

分 类 号：S436.412.1[农业科学—农业昆虫与害虫防治] TP391.41[农业科学—植物保护]