首页|基于ResNet50的水稻病虫害识别

基于ResNet50的水稻病虫害识别

扫码查看
原文链接
  • 国家科技期刊平台
  • NETL
  • NSTL
  • 万方数据
为了准确识别水稻病虫害,收集 8 种水稻病虫害图像和健康水稻图像,构建水稻病虫害数据集.将残差网络ResNet50 用于水稻病虫害识别,在原模型基础上引入了迁移学习和NonLocal注意力机制.实验结果表明,改进模型的准确率、精确率、召回率、F1-score分别达到99.12%、99.31%、99.27%、99.28%,相比于原模型分别提升了2.92%、2.91%、4.05%、3.60%.与模型DenseNet121、Inception V3、ShuffleNet V2、MobileVit-small、ResNext50 相比,改进模型的准确率、精确率、召回率、F1-score至少高出 2%.实验验证了所提模型的有效性,该模型可以准确识别这几种水稻病虫害.
Identification of Rice Pests and Diseases Based on ResNet50
In order to accurately identify rice pests and diseases,8 types of rice pests and diseases images and health rice images are collected,which are used to construct a rice pests and diseases dataset.The Residual Network ResNet50 is used for identification of rice pests and diseases,and Transfer Learning and NonLocal Attention Mechanisms are introduced on the basis of the original model.The experimental results show that the accuracy,precision,recall,and F1-score of the improved model have reached 99.12％,99.31％,99.27％,and 99.28％,respectively,which are 2.92％,2.91％,4.05％,and 3.60％higher than the original model.Compared with models DenseNet121,Inception V3,ShuffleNet V2,MobileVit-small and ResNext50,the improved model has at least 2 percentage points higher on accuracy,precision,recall,and F1-score.The experiment verifies the effectiveness of the proposed model,which can accurately identify these types of rice pests and diseases.

rice pests and diseasesResNet50 modelTransfer LearningNonLocal Attention Mechanism

丁士宁

展开 >

信阳农林学院 信息工程学院,河南 信阳 464000

水稻病虫害 ResNet50模型 迁移学习 NonLocal注意力机制

信阳农林学院青年教师科研基金项目

QN2021057

2024

10.19850/j.cnki.2096-4706.2024.16.027

现代信息科技
广东省电子学会

现代信息科技

ISSN：2096-4706
年,卷(期)：2024.8(16)
  • 11