摘要
槽式杀青理条机是常用的茶叶初加工机械,针对不同的鲜叶状态,固定的杀青理条参数难以保证茶叶在初加工阶段保持最佳的理条状态。文章通过基于机器视觉技术构建的样本质量预测模型,对茶叶鲜叶在槽式理条杀青机内加工的作业效果进行实时的取样拍摄,并提取采集不同理条时间的茶叶的42个外观特征参数(包括9个外观特征、18个颜色特征、15个纹理特征),分别采用决策树模型(DT)、随机森林算法(RF)和支持向量机(SVM)三种分类算法进行实验,三种算法预测的平均准确率为91.5%,其中在测试样本数量较少时决策树算法的识别效果最好,在大量测试样本的训练下支持向量机算法的识别效果最好。运用机器视觉技术可以准确预测茶叶在杀青理条环节的产品质量等级,为茶叶工过程提供参数支持。
Tea slot type fixing and carding machine is one of the commonly used tea primary processing machines.However,it is difficult to ensure that the tea leaves are kept in the best condition for different fresh leaf states during the preliminary processing stage with fixed parameters.In this study,the machine vision technology was used to take real-time samples of tea leaves processed in the slot type carding machine and read 42 appearance features(including 9 appearance features,18 color features and 15 texture features)of tea leaves at different time periods,and adopt decision tree(DT),random forest(RF)and support vector model(SVM)for experiment respectively.The average accuracy of the three models was 0.915,in which the decision tree algorithm had the best recognition effect when the test samples were small and the support vector machine algorithm had the best recognition effect when a large number of test samples were trained.The use of machine vision technology can accurately forecast the operational effect of tea in the fixing and carding process and effectively improve the processing efficiency.
作者
王霄然
吴正敏
钟华
黄洋洋
张雪晨
石震
WANG Xiao-ran;WU Zheng-min;ZHONG Hua;HUANG Yang-yang;ZHANG Xue-chen;SHI Zhen(School of Tea and Food Technology,Anhui Agricultural University,Hefei 230036,China;School of Engineering,Anhui Agricultural University,Hefei 230036,China)
出处
《中国茶叶加工》
2022年第3期40-45,共6页
China Tea Processing
基金
国家级大学生创新训练项目(S202110364001)
安徽省教育厅重点科研项目(KJ2020A0113)
安徽农业大学自然科学青年基金项目(2020zd15)。
关键词
机器视觉技术
数据模型
理条机
外观特征
Machine vision technology
Data model
Carding machine
Appearance characteristics
作者简介
王霄然(2000-),男,安徽合肥人,大学在读,主要从事茶叶智能化加工研究;通讯作者:吴正敏,wuzhengmin@ahau.edu.com。
