欢迎访问《农学学报》,

农学学报 ›› 2020, Vol. 10 ›› Issue (10): 83-90.doi: 10.11923/j.issn.2095-4050.cjas20190700108

所属专题: 水稻

• 农业信息/农业气象 • 上一篇    下一篇

机器视觉在HSV颜色空间下稻瘟病病程分级判定研究

刘永波1(), 雷波1(), 胡亮1, 唐江云1, 曹艳1, 尹亚琳2   

  1. 1四川省农业科学院农业信息与农村经济研究所,成都 610011
    2四川农业大学,成都 611130
  • 收稿日期:2019-07-08 修回日期:2019-09-29 出版日期:2020-10-20 发布日期:2020-11-18
  • 通讯作者: 雷波 E-mail:dylyb618@163.com;689300@sina.com
  • 作者简介:刘永波,男,1988年出生,四川甘洛人,助理研究员,硕士,主要从事计算机视觉技术与农业信息化研究。通信地址:610011 四川省成都市锦江区净居寺路20号附101号信息所,Tel:028-84504208,E-mail: dylyb618@163.com
  • 基金资助:
    四川省科技厅“十三五”农作物及畜禽育种战略研究与云服务平台建设(2016NYZ0054);四川省科技计划项目“基于深度卷积神经网络的玉米病害智能识别与分级鉴定研究”(2018JY0631)

The Grading Determination of Rice Blast: HSV Color Space Method Based on Machine Vision

Liu Yongbo1(), Lei Bo1(), Hu Liang1, Tang Jiangyun1, Cao Yan1, Yin Yalin2   

  1. 1Institute of Agricultural Information and Rural Economy, Sichuan Academy of Agricultural Sciences, Chengdu 610011, Sichuan, China
    2Sichuan Agricultural University, Chengdu 611130, Sichuan, China
  • Received:2019-07-08 Revised:2019-09-29 Online:2020-10-20 Published:2020-11-18
  • Contact: Lei Bo E-mail:dylyb618@163.com;689300@sina.com

摘要:

该研究旨在开发基于机器视觉技术的稻瘟病病程分级系统,实现对稻瘟病病程分级准确、客观的判定。提出基于GrabCut、高斯滤波、OTSU二值化、颜色空间转换、阈值切割等处理的稻瘟病分级判定算法模型,该算法模型利用OpenCV与python语言实现,以反向阈值切割为核心策略分离叶片与病斑,再以循环遍历模式统计像素点得出病斑面积占比,实现对稻瘟病的快速、精确分级。试验结果表明,该算法模型与专业研究人员人工判定的结果匹配度达95.77%,相对于人工判定,具备更高的稳定性和客观性。目前对稻瘟病病程分级主要依赖研究人员通过经验判定,客观、准确的判定病程对防治稻瘟病具有重要意义。该系统以手机APP为图像采集端口,不依赖其他仪器和设备,通过手机拍照即可实时获得稻瘟病精确的分级结果,降低了研究门槛,提高了科研工作的效率。

关键词: 水稻病害, 稻瘟病, 图像处理, 机器视觉, GrabCut, 病程分级

Abstract:

The purpose is to develop a grading system for the disease course of rice blast based on machine vision technology, so as to realize accurate and objective classification of the disease course of rice blast. Based on GrabCut, gaussian filter, OTSU binarization, color space conversion, threshold cutting, etc., an algorithm model of rice blast classification judgment was proposed. The algorithm model is implemented by OpenCV and python, and the reverse threshold cutting is taken as the core strategy to separate the leaves from the diseased spots, and then the percentage of diseased spots is calculated by the cyclic traversal model to realize the rapid and accurate classification of rice blast. The result of the algorithm model matches the manual judgment by professional researchers up to 95.77%. Compared with the manual judgment, the algorithm model has higher stability and objectivity. At present, the classification of disease course of rice blast mainly depends on the experience judgment of researchers. Objective and accurate determination of disease course is of great significance for the prevention and treatment of rice blast. Mobile phone APP is used as image acquisition port, which does not rely on other instruments and equipment and could obtain accurate grading results of rice blast in real time, reduce the threshold of research and improve the efficiency of scientific research.

Key words: Rice Diseases, Rice Blast, Image Processing, Machine Vision, GrabCut, Disease Course Grading

中图分类号: