摘要
针对人工烟叶分级的差异性、分级结果不稳定、合格率低等问题,提出了一种基于卷积神经网络(convolutional neural networks,CNN)的烟叶分级检测算法.首先,选取inception V3模型,结合卷积层与网络层完成迁移学习,运用多分类的模型选取交叉熵作为损失函数进行运算;同时,与极值点跳跃算法相结合,对采集的烟叶进行识别分析;仿真实验验证了该方法的有效性,得到了较好的识别效果.
For the differences in the classification of artificial tobacco leaves,unstable classification results,and low qualification rates,this paper proposes a tobacco leaf grading detection algorithm based on Convolutional Neural Networks(CNN).First,select the inception V3 model,combine the convolution layer with the network layer to complete migration learning,The multi-class model is used to select the cross entropy as the loss function for calculation;At the same time,the algorithm proposed in this paper is combined with the extreme point jump algorithm to identify and analyze the collected tobacco leaves.The simulation results verify the effectiveness of the proposed method and obtain a good recognition result.
作者
王士鑫
云利军
叶志霞
王一博
WANG Shi-xin;YUN Li-jun;YE Zhi-xia;WANG Yi-bo(School of Information,Yunnan Normal University,Kunming 650500,China;Yunnan Key Laboratory of Opto-electronic Information Technology,Kunming 650500,China)
出处
《云南民族大学学报(自然科学版)》
CAS
2020年第1期65-69,共5页
Journal of Yunnan Minzu University:Natural Sciences Edition
基金
云南省应用基础研究计划重点项目(2018FA033).
关键词
烟叶分级
卷积神经网络
迁移学习
tobacco leaf grading
convolutional neural network
migration learning