基于GoogLeNet改进模型的苹果叶病诊断系统设计  

Design of apple leaf disease diagnosis system based on GoogLeNet improved model

作　　者：宋晨勇 白皓然[1] 孙伟浩 马皓冉 Song Chenyong;Bai Haoran;Sun Weihao;Ma Haoran(College of Mechanical and Electrical Engineering,Qingdao Agricultural University,Qingdao,266109,China)

机构地区：[1]青岛农业大学机电工程学院,山东青岛266109

出　　处：《中国农机化学报》2021年第7期148-155,共8页Journal of Chinese Agricultural Mechanization

基　　金：山东省农机装备研发创新计划项目(2018YZ002);国家级大学生创新创业训练计划项目(S202010435013);广东省重点领域研发计划(2018B020241003)。

摘　　要：针对苹果病害自动识别精度低、高延迟问题,以苹果锈病和斑点落叶病作为研究对象,基于卷积神经网络建立苹果病害诊断系统。采用旋转、锐化、模糊等数据增强方法扩展数据集,使用数据标准化方法优化模型的数据输入,在算法模型上,在GooLeNet模型的基础上减少inception模块上的数量,并优化模型前段的网络结构以捕获更多的特征信息,最后系统依据建立的算法模型和Tornado框架构建病害识别网络视图。试验结果表明:GoogLeNet改进模型的精准率提高2.1%,达到98.4%,模型参数仅为原模型的17.5%,实现苹果病害实时、便利的图像自动识别,为苹果病害防治防控提供可靠的参考依据。In order to overcome the shortcomings of low accuracy and high delay in the automatic identification of apple diseases,this paper took apple rust and spotted leaf disease as the research objects and established an apple disease diagnosis system based on Convolutional Neural Network.This research used data enhancement methods such as rotation,sharpening,and blurring to expand the data set and used data standardization methods to optimize the data input of the model.On the algorithm model,this research reduced the number of inception modules based on the GooLeNet model and optimized the network structure in the front part of the original model,which was used to capture more characteristic information.Finally,the system built a disease recognition network view based on the established algorithm model and Tornado framework.The experimental results showed that the accuracy of the improved Googlenet model was increased by 2.1%to 98.4%,and the model parameters were only 17.5%of the original model,which realized real-time and convenient automatic image recognition of apple diseases.This paper can provide a reliable reference for the prevention and control of apple diseases.

关 键 词：苹果叶病识别 数据增强 GoogLeNet改进模型 Tornado框架 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]