欢迎访问《茶叶科学》,今天是

茶叶科学 ›› 2023, Vol. 43 ›› Issue (5): 691-702.doi: 10.13305/j.cnki.jts.2023.05.007

• 研究报告 • 上一篇    下一篇

基于图像处理技术的茶树新梢识别和叶面积计算的探索研究

吕丹瑜1, 金子晶2, 陆璐1, 何卫中3, 疏再发3, 邵静娜3, 叶俭慧1,*, 梁月荣1   

  1. 1.浙江大学茶叶研究所,浙江 杭州 310058;
    2.浙江省农业技术推广中心,浙江 杭州 310000;
    3.丽水市农林科学研究院,浙江 丽水 323000
  • 收稿日期:2023-05-29 修回日期:2023-08-23 出版日期:2023-10-15 发布日期:2023-11-06
  • 通讯作者: *jianhuiye@zju.edu.cn
  • 作者简介:吕丹瑜,女,主要从事茶叶生产数字化品控技术方面研究,3190101348@zju.edu.cn。
  • 基金资助:
    浙江省农业重大技术协同项目(2020XTTGCY04)、浙江省农业(茶树)新品种选育重大科技专项(2021C02067-5-1)

Exploratory Study on the Image Processing Technology-based Tea Shoot Identification and Leaf Area Calculation

LÜ Danyu1, JIN Zijing2, LU Lu1, HE Weizhong3, SHU Zaifa3, SHAO Jingna3, YE Jianhui1,*, LIANG Yuerong1   

  1. 1. Tea Research Institute, Zhejiang University, Hangzhou 310058, China;
    2. Zhejiang Agricultural Technical Extension Center, Hangzhou 310000, China;
    3. Lishui Institute of Agriculture and Forestry Sciences, Lishui 323000, China
  • Received:2023-05-29 Revised:2023-08-23 Online:2023-10-15 Published:2023-11-06

摘要: 基于田间采集的大量茶树春梢生育图片,借助计算机视觉技术,利用目标检测算法YOLOv5构建茶树新梢不同生育阶段的识别模型,测试结果表明该模型具有较高的检测精度。进一步探究了Image-J软件处理以及基于Gray值、RGB值、HSV值的阈值切割图像处理方法在茶叶面积处理方面的应用,比较了不同方法的准确度和运行效率。结果表明,基于HSV阈值切割法的茶树叶片面积算法准确率在94%以上,表现优于RGB阈值切割法。研究结果为开发茶树新梢生育进度智能识别模型和叶片性状信息提取算法提供了技术支撑,为采茶机械的茶芽自动识别模块的研发提供了理论基础。

关键词: 计算机图像处理技术, YOLO, 茶树新梢, 生育阶段, 叶面积, 模型构建

Abstract: In this study, based on the picture collection of tea shoot growth in the field, we used deep learning target detection algorithm YOLOv5 to construct a model for identifying different growth stages of tea shoots, and the testing results indicate that the model had high accuracy. Furthermore, the Image-J software and the image processing methods of threshold cutting based on Gray, RGB and HSV values were applied to process tea leaf area, and the accuracy and efficiency of different methods were compared. The results show that the accuracy of HSV-based algorithm system of cutting tea leaves and automatically calculating tea leaf area was over 94%, which had better performance than RGB-based algorithm system. The research results provide technical support for the intelligent recognition model of tea growth state and information extraction algorithm of leaf traits, and also build a theoretical basis for the development of tea bud automatic recognition module of tea plucking machinery.

Key words: computer image processing technology, YOLO, tea shoot, growth stage, leaf area, model building

中图分类号: