基于改进ShuffleNetV2的荔枝叶片病虫害图像识别  

Image recognition of leaf pests and diseases based on improved ShuffleNet V2 in litchi

作　　者：谢家兴 陈斌瀚 彭家骏 何培华 景庭威 孙道宗 高鹏 王卫星 郑代德 李君 XIE Jiaxing;CHEN Binhan;PENG Jiajun;HE Peihua;JING Tingwei;SUN Daozong;GAOPeng;WANG Weixing;ZHENG Daide;LI Jun(College of Electronic Engineering(College of Artificial Intelligence)of South China Agricultural University,Guangzhou 510642,Guangdong,China;Maoming Branch,Guangdong Laboratory for Lingnan Modern Agriculture,Maoming 525000,Guangdong,China;Engineering Research Center for Monitoring Agricultural Information of Guangdong Province,Guangzhou 510642,Guangdong,China;College of Engineering of South China Agricultural University,Guangzhou 510642,Guangdong,China)

机构地区：[1]华南农业大学电子工程学院(人工智能学院),广州510642 [2]岭南现代农业科学与技术广东省实验室茂名分中心,广东茂名525000 [3]广东省农情信息监测工程技术研究中心,广州510642 [4]华南农业大学工程学院,广州510642

出　　处：《果树学报》2023年第5期1024-1035,共12页Journal of Fruit Science

基　　金：华南农业大学新农村发展研究院农业科技合作共建项目(2021XNYNYKJHZGJ032);茂名实验室自主科研项目(2021ZZ002);广东省现代农业产业技术体系创新团队建设专项(2022KJ108);岭南现代农业实验室科研项目(NT2021009);广东省乡村振兴战略专项(农业科技能力提升)(TS-1-4);财政部和农业农村部:国家现代农业产业技术体系(CARS-32-14);广东省科技创新战略专项资金(“攀登计划”专项资金)项目(pdjh2021b0077,pdjh2021b0077);大学生创新创业训练计划项目。

摘　　要：【目的】探索及时、准确识别危害荔枝叶片病虫害的方法,做好防护和治理工作。【方法】以常见荔枝叶片病虫害图像为研究对象,针对荔枝病虫害图像的病斑分布区域广、面积大小不一等特点,对ShuffleNet V2模型进行相应改进。首先,采集荔枝5类叶片病虫害制作数据集,采用数据增强操作构建更加丰富的数据集;其次,在网络特征提取模块采用混合空洞卷积,避免传统卷积在下采样过程中造成图像分辨率降低及信息丢失;然后,通过嵌入轻量型通道注意力模块ECA(efficient channel attention),增进特征图之间相互依赖关系。此外,删减模型中不必要的层数和通道数,降低模型的参数量及运算量。【结果】研究所改进模型在5类荔枝病虫害图像(毛毡病、炭疽病、煤烟病、叶瘿蚊、藻斑病)中达到了99.04%的识别准确率,比原网络ShuffleNet V2高出2.55%。相较于经典网络AlexNet、ResNet-18、DenseNet和MobileNet V2等,改进模型不仅有着更高的识别准确率,并且改进后模型参数量仅为0.059×10^(6),为原模型的4.92%,浮点运算量仅为0.183×10^(9)。【结论】研究结果适合部署在移动终端等嵌入式资源受限设备上,有助于实现对作物病虫害实时、准确地识别。【Objectives】Litchi suffers from many kinds of pests and diseases.Therefore,it is necessary to invest enough energy and funds to control them to ensure the normal growth of litchi.At present,the leaf pest and disease identification of litchi is a problem urgently to be solved.In order to explore ways to identify litchi leaf pests and diseases in a timely and accurate manner,the experiment was undertaken,so as to take preventive and control measures in a timely manner.In this study,common leaf pest and disease images were taken as the research objects,and the ShuffleNet V2 model was improved accordingly for the difficulty to accurately identify if the types of pests and diseases on litchi leaves had the characteristics of large distribution area and different size of lesions.【Methods】First,five types of litchi leaf pests and diseases (Algal leaf spot, Aceria litchi, Sooty mold, Anthraconse and Dasineura sp.)were collected as the data set for the model test. In order to improve the robustness of the model, theoriginal data were augmented by methods such as flipping, cropping, adding noise and changing contrastto obtain a more abundant data set. Second, hybrid dilated convolution was used in the network featureextraction module to obtain a larger receptive field to avoid the loss of information during thedownsampling process, and to eliminate the local information loss caused by the use of ordinary dilatedconvolution stacking. The bigger the receptive field, the larger the range of the corresponding originalimage, which means that it contains more global and higher semantic level features. In terms of imageclassification, the region of interest is often distributed in multiple areas of the image, and more globalinformation and higher-level feature information are needed to better identify the target. Third, the attentionmechanism can better aggregate the feature information of the target to be recognized by the networkmodel and reduce the influence of irrelevant background. By embedding a lightweight channel

关 键 词：荔枝 叶斑症状 图像识别 ShuffleNetV2模型 模型参数 

分 类 号：S667.1[农业科学—果树学]