摘要
智慧农业是未来农业的发展方向。治理农作物病害最关键的问题是精确的判断病害类型。为了进一步提高农作物病害识别准确率，提出一种改进ConvNeXt的农作物病害识别方法，为了提升网络的特征提取能力，将ConvNeXt基本模块中加入无参注意力模块SimAM；为了加强通道的特征，在原有的网络结构上增加ECA（Efficient Channel Attention）注意力模块；在公开数据集PlantVillage上进行试验。结果表明：与原始的ConvNeXt模型相比，改进后的方法在不增加参数量的同时，识别准确率达到99.36%，比原模型提高了2.02个百分点，为农作物的自动化识别提供参考。
关键词： 智慧农业；病害识别；ConvNeXt；SimAM；注意力机制
Abstract
Smart agriculture is the development direction of future agriculture. The most crucial issue in managing crop diseases is to accurately determine the type of disease. In order to further improve the accuracy of crop disease recognition, an improved ConvNeXt method for crop disease recognition is proposed. To enhance the feature extraction ability of the network, a parameter free attention module SimAM is added to the network structure of ConvNeXt; In order to enhance the characteristics of channels, an ECA attention module is added to the basic module of the model; Conduct experiments on the publicly available dataset PlantVillage. The results show that compared with the original ConvNeXt model, the improved method achieves a recognition accuracy of 99.36% without increasing the number of parameters, which is 2.02 percentage points higher than the original model, providing a reference for automated crop recognition.
Key words： Smart agriculture; Disease identification; ConvNeXt; SimAM; Attention mechanism
参考文献 References
[1] N. Krishnamoorthy, L.V. N. Prasad, C. P. Kumar, B. Subedi, H. B. Abraha et al, "Rice leaf diseases prediction using deep neural networks with transfer learning," Environmental Research, vol. 198, no.11, pp. 111275, 2021.
[2] 姚建斌,张英娜,刘建华.基于卷积神经网络和迁移学习的小麦病虫害识别[J].华北水利水电大学学报(自然科学版),2022,43(02):102-108.
[3] D. Hughes and M Salathé, "An open access repository of images on plant health to enable the development of mobile disease diagnostics," arXiv preprint arXiv:1511.08060, 2015.
[4] Yang L, Zhang R Y, Li L, et al. Simam: A simple, parameter-free attention module for convolutional neural networks[C]//International conference on machine learning. PMLR, 2021: 11863-11874.
[5] 朱传军,刘荣光,成佳闻等.基于SimAM模块与ResNet34网络的混合缺陷检测模型[J].现代制造工程,2023,No.509(02):1-9.
[6] Q. Wang, B. Wu, P. Zhu, P. Li and Q. Hu, “ECA-Net: Efficient channel attention for deep convolutional neural networks,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, SEA, USA, pp. 13-19, 2020.