融合分组注意力机制的水稻病虫害图像识别算法  

Image recognition algorithm of rice diseases and insect pests based on shuffle attention mechanism

作　　者：高雨亮 徐向英[1] 章永龙[1] 顾逸枫 张力峰 李斌[1] GAO Yuliang;XU Xiangying;ZHANG Yonglong;GU Yifeng;ZHANG Lifeng;LI Bin(School of Information Engineering,Yangzhou University,Yangzhou 225127,China)

机构地区：[1]扬州大学信息工程学院,江苏扬州225127

出　　处：《扬州大学学报（自然科学版）》2021年第6期53-57,共5页Journal of Yangzhou University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61472344,61611540347);教育部农业与农产品安全国际合作联合实验室课题(JILAR-KF202007).

摘　　要：针对传统的水稻病虫害图像识别算法缺乏注意力权重、准确率低下等问题,提出一种基于深度学习的水稻病虫害识别算法.在Inception-ResNetV1卷积模块引入分组注意力(shuffle attention,SA),使得注意力模块参数量大幅降低并提升图像识别的准确度.在自制的水稻病虫害数据集中,病虫害图像识别的准确度高达95.55%,与传统模型ResNet50和Inception-ResNetV1相比,模型参数量分别下降62.7%和32.5%,准确度分别上升1.8%和1.21%.This paper presents an identification algorithm of rice diseases and insect pests based on convolutional neural network.Aiming at the problems of lacking of attention weight and low accuracy of traditional algorithm,the traditional convolution neural network Inception-ResNetV1 and the shuffle attention module are combined.It achieves the effects of greatly reducing parameters and obviously improving accuracy.Compared with traditional models ResNet50 and Inception-ResNetV1,the highest accuracy reaches 95.5%the parameters in the self-made data set are decreased by 62.7%and 32.5%,and the accuracy is increased by 1.8%and 1.21%,respectively.

关 键 词：水稻病虫害识别 Inception-ResNetV1 分组注意力 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] S41-30[自动化与计算机技术—计算机科学与技术]