基于深度学习的茶叶种类识别系统  

The Recognition System of Tea Based on Deep Learning

作　　者：李国志 冯泽霖 方梦瑞 吕军 

机构地区：[1]浙江理工大学信息学院,浙江 杭州

出　　处：《软件工程与应用》2021年第4期568-575,共8页Software Engineering and Applications

摘　　要：针对茶叶种类多、炒青茶特征不显著,难以快速辨别炒青茶种类的问题,开发了一款基于深度学习的茶叶种类识别系统。采集西湖龙井、安吉白茶、黄山毛峰、建德黄金芽和绩溪金山时雨5种炒青茶图像1048幅,利用图像锐化、水平镜像、垂直镜像、灰度处理和图像旋转等方法进行数据增强,并按照8:1:1的比例划分训练集、验证集和测试集;建立茶叶种类及相关信息数据库,训练了基于VGG16的茶叶种类识别模型,并将茶叶数据库和识别模型部署在服务器;开发了基于Android的茶叶种类识别APP,具有地图定位、茶叶信息搜索、茶叶智能识别和用户信息等模块。经系统测试,茶叶测试集平均识别率为98.1%,系统平均响应时间为12 s,该系统为快速识别茶叶种类提供理论参考和使用工具,对稳定市场秩序和维护消费者权益具有实际应用价值。To solve the problem of the high number of tea, the non-significant characteristics of roasted tea and the difficulty in quickly identifying the type of roasted tea, a tea recognition system based on deep learning has been developed. 1048 images of five kinds of roasted tea were collected;image sharpening, horizontal mirroring, vertical mirroring, grayscale processing, and image rotation were used for data Enhance;the training set, validation set and test set according to the ratio of 8:1:1 was divided;the database of tea and tea identification model based on VGG16 were deployed on the server;an Android-based tea identification APP with modules such as map positioning, tea information search, tea intelligent identification and user information was developed. According to the results, the average recognition rate of the tea test set is 98.1%, and the average response time of the system is 12 s. The system provides theoretical references and tools for quickly identifying tea, and has practical application value for stabilizing market order and safeguarding consumer rights.

关 键 词：茶叶 自动识别 深度学习 移动终端 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]