基于深度学习的在线烟叶等级判定研究

Online Tobacco Leaf Grade Determination Research Based on Deep Learning

对烟叶进行等级判定可以合理利用烟叶资源,提高卷烟产品质量,对实现经济利益最大化有重要意义。提出一种基于深度学习的在线烟叶等级判定方法,该方法采用ResNeXt为基础网络,在残差结构中嵌入SE模块以增强重要通道的信息,并引入FPN+PAN结构将网络浅层细节特征和高层语义特征进行融合,以实现多尺度特征表达。测试结果表明,该方法烟叶等级判定的准确率达到92.8%,因此该方法对烟叶等级具备良好识别的能力,可适用实际生产。

The grading judgment of tobacco leaves can rationally utilize tobacco resources,improve the quality of cigarette products,and is of great significance to maximize economic benefits.An online tobacco leaf grade determination method based on deep learning was proposed,which used ResNeXt as the basic network,embedded the SE module in the residual structure to enhance the information of important channels,and introduced the FCN+PAN structure to fuse the shallow detail features and high-level semantic features of the network to achieve multi-scale feature expression.The test results showed that the accuracy of the tobacco grade determination of the method reached 92.8%,which showed that the method had the ability to identify the tobacco grade well and could be applied to actual production.