浙江农业学报 ›› 2023, Vol. 35 ›› Issue (2): 445-454.DOI: 10.3969/j.issn.1004-1524.2023.02.22

• 生物系统工程 • 上一篇    下一篇

基于YOLOv5的小麦种子发芽检测方法研究

白卫卫1,2(), 赵雪妮1, 罗斌2, 赵薇1,2, 黄硕3, 张晗2,*()   

  1. 1.陕西科技大学,陕西 西安 710016
    2.北京市农林科学院 智能装备技术研究中心,北京 100097
    3.北京市农林科学院 信息技术研究中心,北京 100097
  • 收稿日期:2020-04-12 出版日期:2023-02-25 发布日期:2023-03-14
  • 通讯作者: 张晗
  • 作者简介:白卫卫(1995—),男,陕西宝鸡人,硕士研究生,研究方向为智能检测与自动化控制技术。E-mail:985782044@q.com
  • 基金资助:
    国家重点研发计划(2017YFD701205);北京市农林科学院青年基金(QNJ1202104);北京市农林科学院2022年度科研创新平台建设项目(PT2022-34)

Study of YOLOv5-based germination detection method for wheat seeds

BAI Weiwei1,2(), ZHAO Xueni1, LUO Bin2, ZHAO Wei1,2, HUANG Shuo3, ZHANG Han2,*()   

  1. 1. Shaanxi University of Science and Technology, Xi’an 710016, China
    2. Research Center of Intelligent Equipment, Beijing Academy of Agriculture and Forestry Sciences, Beijing 100097, China
    3. Research Center of Information Technology, Beijing Academy of Agriculture and Forestry Sciences, Beijing 100097, China
  • Received:2020-04-12 Online:2023-02-25 Published:2023-03-14
  • Contact: ZHANG Han

摘要:

种子发芽试验是检验作物品质的重要环节。为提高种子发芽检测效率,实现种子发芽检测自动化,以小麦为研究对象,通过机器视觉技术结合深度学习方法,构建基于YOLOv5的种子发芽判别的模型,在此基础上通过小麦7 d发芽试验图像组合分析,设计一套基于YOLOv5的种子发芽检测改进判别方法(DB-YOLOv5),实现对小麦种子发芽率、发芽势、发芽指数、平均发芽天数的快速检测,并开展检测试验。结果表明,YOLOv5模型对小麦种子发芽判别精确率为92.5%,DB-YOLOv5模型对小麦种子发芽判别精确率为98.5%,发芽势、发芽指数、平均发芽天数与人工检测误差为0.5%、2.39、0.1 d。上述结果表明,DB-YOLOv5模型可实现对小麦种子发芽率、发芽势、发芽指数、平均发芽天数的快速检测,为农作物种子发芽快速检测提供参考。

关键词: 小麦种子, 发芽检测, 深度学习, YOLOv5

Abstract:

Seed germination test is an important part of testing the quality of crops. In order to improve the efficiency of seed germination detection and realize the automation of seed germination detection, taking wheat as the research object, a model based on YOLOv5 seed germination discrimination was constructed by machine vision technology combined with deep learning methods. Based on this, a set of improved discriminative methods for seed germination detection based on YOLOv5 was designed through the combined analysis of wheat germination test images in 7 days. The rapid detection of wheat seed germination rate, germination potential, germination index and average germination days was realized and the detection experiment was carried out. The results showed that the YOLOv5 model had 92.5% accuracy for wheat seed germination. By using the improved discrimination method of seed germination detection based on YOLOv5, the accuracy of seed germination discrimination was 98.5%, and the errors of germination potential, germination index, and average germination days were 0.5%, 2.39, and 0.1 d compared with manual detection. The improved discrimination method based on YOLOv5 seed germination detection proposed in this study could realize the rapid detection of seed germination rate, germination potential,germination index and average germination days, and provide a reference for the rapid detection of crop seed germination.

Key words: wheat seed, germination detection, deep learning, YOLOv5

中图分类号: