摘要
玉米是我国重要的农作物之一,其叶片的生长状态直接影响玉米的产量。玉米叶片病虫害种类多达数十种,依赖人工识别不仅需要专业的知识,还消耗大量的精力,且由于人眼生理结构限制,通常会有较高的识别误差。为了提高对玉米叶片病虫害识别的准确率,本文应用卷积神经网络搭建识别模型,对玉米病虫害叶片进行分类,取得了较高的识别准确率,针对训练集的平均精度达到96.25%,测试集平均精度达到93.74%;同时文中探讨了玉米叶片尺寸的变化对模型识别率的影响,实验结果显示,相同模型深度、参数情况下,尺寸越大越不利于提升模型的精度,同时消耗了大量的时间,平均精度也会随之降低。
Corn is one of the important crops in China.The growth state of its leaves directly affects the health and yield of corn.There are dozens of species of corn leaf diseases and insect pests.Relying on manual identification not only requires professional knowledge,but also consumes a lot of energy.Due to the limitation of the physiological structure of the human eye,there is usually a high recognition error.In order to improve the accuracy of the recognition of corn leaf diseases and insect pests,this paper uses a convolutional neural network to build a recognition model to classify corn leaf diseases and insect pests.It has high recognition accuracy,the average accuracy in the training set reaches 96.25%,and the average accuracy in the test set reaches 93.74%.At the same time,the influence of the change of corn leaf size on the recognition rate of the model is discussed.The experimental results show that the same model depth and parameters,a larger size is not conducive to improvement of accuracy of the model,and it will consume a lot of time,and the average accuracy will also decrease.
作者
张立
林祥锐
张钰
郭春阳
ZHANG Li;LIN Xiangrui;ZHANG Yu;GUO Chunyang(Guangdong Baiyun University,Guangzhou 510450,China)
出处
《智能物联技术》
2022年第2期28-31,共4页
Technology of Io T& AI
基金
广东白云学院校级课题(2021BYKY26)。
关键词
深度学习
玉米病虫害
分类识别
卷积神经网络
deep learning
corn pests and diseases
classification and identification
convolutional neural network
