| 基于多特征融合的茶叶鲜叶等级识别的方法研究 |
| |
| 引用本文： | 张金炎,曹成茂,李文宝,王二锐,孙燕,刘光宗.基于多特征融合的茶叶鲜叶等级识别的方法研究[J].安徽农业大学学报,2021,48(3):480-487. |
| |
| 作者姓名： | 张金炎 曹成茂 李文宝 王二锐 孙燕 刘光宗 |
| |
| 作者单位： | 安徽农业大学工学院,合肥230036;安徽省智能农机装备工程实验室,合肥230036;农业农村部南方农业装备科学观测实验站,合肥230036 |
| |
| 基金项目： | 安徽省科技重大专项（18030701195）和安徽农业大学2020年度研究生创新基金项目（2020ysj-74）共同资助。 |
| |
| 摘 要： | 茶叶鲜叶等级直接影响优质绿茶成品的等级,如果在鲜叶阶段就茶叶的芽叶数量进行等级识别,并将不同等级鲜叶分离出来,制作不同等级的绿茶成品,从一定程度上解决了优质绿茶鲜叶采摘环节的难题.提出基于茶叶形态、纹理和HOG特征的鲜叶分级方法,采集鲜叶样本图片,对样本图片进行预处理操作,再提取鲜叶形态和纹理特征等特征参数,建立机器学习模型支持向量机、随机森林和线性判别法K-最近邻对新鲜茶叶样本进行分类,得到各等级的茶叶识别结果.试验结果表明,单独使用一种特征分类效果不佳,也不符合茶叶本身的复杂性.将多种特征融合有更好的分类效果;3种算法中,随机森林算法有较高的优越性,准确率达97.06％.该研究提取的多特征参数和分类模型,为实际鲜叶的生产加工等级识别提供参考.
|
| 关 键 词： | 特征融合 随机森林 机器学习 鲜叶等级识别 HOG特征 |
Study on the method of recognition of fresh leaf grade of tea based on multi-feature fusion |
| |
| ZHANG Jinyang,CAO Chengmao,LI Wenbao,WANG Errui,SUN YanLIU Guangzong.Study on the method of recognition of fresh leaf grade of tea based on multi-feature fusion[J].Journal of Anhui Agricultural University,2021,48(3):480-487. |
| |
| Authors: | ZHANG Jinyang CAO Chengmao LI Wenbao WANG Errui SUN YanLIU Guangzong |
| |
| Institution: | School of Engineering, Anhui Agricultural University, Hefei 230036; Anhui Province Engineering Laboratory of Intelligent Agricultural Machinery Equipment, Hefei 230036; Scientific Observing and Experimental Station of Agriculture Equipment for the Southern China Ministry of Agricultural, Hefei 230036 |
| |
| Abstract: | Tea fresh leaf grade directly affects the grade of high-quality green tea finished products, if the number of buds and leaves of tea in the fresh leaf stage for grade identification, and different grades of fresh leaves separated, the production of different levels of green tea finished products, to a certain extent to solve the high-quality green tea fresh leaf picking link of the problem. In this paper, the fresh leaf grading method based on tea morphological, texture and HOG features is proposed, the fresh leaf sample picture is collected, the sample picture is pre-processed, the features parameters such as fresh leaf morphological and texture feature situ, and the machine learning model supports the vector machine, the random forest and the linear discrimination method K-nearest neighbor to classify the fresh tea sample, and gets the tea recognition results of each grade. The experimental results show that the use of a feature classification alone is not effective and does not conform to the complexity of the tea itself. The fusion of various features has a better classification effect, and among the three algorithms, the random forest algorithm has higher advantages, with an accuracy rate of 97.06%. The multi-featured parameters and classification models extracted in this study provide reference for the identification of the production and processing level of actual fresh leaves. |
| |
| Keywords: | feature fusion random forest machine learning fresh leaf grade recognition HOG features |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《安徽农业大学学报》浏览原始摘要信息 |
| 点击此处可从《安徽农业大学学报》下载免费的PDF全文 |
|
