引用本文:李建华,郝 炘※,牛明雷,王俊伟,李平安,杨立国.基于卷积神经网络的农作物病害识别[J].中国农业信息,2019,31(3):39-47
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 1130次   下载 494 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于卷积神经网络的农作物病害识别
李建华1, 郝 炘※2, 牛明雷3, 王俊伟4, 李平安5, 杨立国6
1.天津市宝坻区朝霞街道办事处,天津301800;2.天津农垦渤海农业集团有限公司,天津301823;3.农业农村部工程建设服务中心,北京100081;4.北京市植物保护站,北京100029;5.湖南省益阳市桃江县农业局,益阳413499;6.内蒙古自治区植保植检站,呼和浩特010010
摘要:
【目的】农作物生长过程中,作物产量会受到各种病害影响,实现自动精准地识别农作物病害以及病害程度的测定是农作物病害防治的关键。【方法】文章设计了一种基于卷积神经网络的农作物病害的识别方法并建立了农作物病害识别模型,模型利用10 种作物中常见的59 种病害类型的叶片图像数据集进行训练,并对模型的训练过程和训练结果进行评估。【结果】(1)农作物病害识别模型对59 种病害类型的总识别精度达到0.83,部分类别的识别率高于0.9;(2)当训练的迭代次数增加到50 轮以上时,农作物病害识别模型的性能不再提升,此时数据集图像的数量对模型性能的影响较大。【结论】实验证明,利用卷积神经网络进行农作物病害识别具有较高的可行性和准确性,为农作物病害的防治打下基础。
关键词:  卷积神经网络  农作物病害  图像识别
DOI:10.12105/j.issn.1672-0423.20190304
分类号:
基金项目:
Crop disease identification based on convolutional neural network
Li Jianhua1, Hao Xin※2, Niu Minglei3, Wang Junwei4, Li Pingan5, Yang Liguo6
1.Zhaoxia Street Office,Baodi District,Tianjin 301800,China;2.Tianjin Agricultural Reclamation Bohai Agricultural Group Co.,Ltd,Tianjin 301823,China;3.Construction Service Center,Ministry of Agriculture and Rural Affairs,Beijing 100081,China;4.Beijing Plant Protection Station,Beijing 100029,China;5.Agricultural Bureau of Taojiang County,Yiyang 413499,China;6.Plant Protection and Plant Inspection Station of Inner Mongolia Autonomous Region,Hohhot 010010,China
Abstract:
[Purpose]During crop growth,yield is affected by various diseases. The realization of automatic and accurate identification of crop diseases and the determination of disease degree is key to the prevention and control of crop diseases.[ Method]In this paper we design a crop disease identification method based on convolutional neural network and establishes a crop disease identification model. The model was trained by using image datasets of 24 diseased leaves in 10 crops,and the training process and training results of the model were evaluated. [Result]The identification model of crop diseases has a total recognition accuracy of about 0.83 for 59 categories,and the recognition rate of some categories is higher than 0. 9. When the number of iterations of training increases to more than 50 rounds,the performance of the crop disease identification model is no longer improved. At this time,the number of dataset images has a greater impact on the model performance.[ Conclusion]Experiments show that the use of convolutional neural networks for crop disease identification has high feasibility and accuracy,laying a foundation for the prevention and control of crop diseases.
Key words:  convolutional neural network  crop diseases  image identification