Journal of Agricultural Science and Technology ›› 2022, Vol. 24 ›› Issue (5): 111-118.DOI: 10.13304/j.nykjdb.2021.0935

• INTELLIGENT AGRICULTURE & AGRICULTURAL MACHINERY • Previous Articles    

Research and Application of Intelligent Recognition Method of Peach Tree Diseases Based on AI

Jianwei WU1,2,3(), Jie HUANG1, Xiaofei XIONG1,2,3, Han GAO3, Xiangyang QIN1()   

  1. 1.Beijing Academy of Agriculture and Forestry Sciences,Beijing 100097,China
    2.China National Engineering Research Center for Information Technology in Agriculture,Beijing 100097,China
    3.Beijing PAIDE Science and Technology Development Co. ,Ltd.,Beijing 100097,China
  • Received:2021-11-03 Accepted:2022-01-18 Online:2022-05-15 Published:2022-06-06
  • Contact: Xiangyang QIN

基于AI的桃树病害智能识别方法研究与应用

吴建伟1,2,3(), 黄杰1, 熊晓菲1,2,3, 高晗3, 秦向阳1()   

  1. 1.北京市农林科学院, 北京 100097
    2.国家农业信息化工程技术研究中心, 北京 100097
    3.北京派得伟业科技发展有限公司, 北京 100097
  • 通讯作者: 秦向阳
  • 作者简介:吴建伟 E-mail:wujw@nercita.org.cn
  • 基金资助:
    北京市科技计划项目(Z211100004621004);北京市农林科学院项目(2021109);北京市乡村振兴科技项目(2022)

Abstract:

In order to solve the problems of low efficiency, high cost and low accuracy of traditional methods of manually identifying peach tree diseases, an intelligent recognition method of peach tree diseases based on AI deep learning was proposed. Using and fine-tuning the DenseNet-169 classification model pre-trained by ImageNet, data preprocessing and model training were performed on the image of 11 common diseases of peach trees, then the web terminal was built to integrate and develop a software system for intelligent recognition of peach tree diseases. The average recognition rate of these 11 peach tree diseases was over 91% by this method. Using image recognition technology to automatically identify peach tree diseases, combined with modern science and technology such as image processing, deep learning, data mining and analysis, the intelligent diagnosis and prevention suggestions for peach tree diseases were realized. This proposed method could reduce labor costs, simplify operations, and improve recognition efficiency, so it was conducive to timely diagnosis and decision-making for prevention and treatment of diseases. Therefore, this research had important significance and application value for promoting the intelligent management of orchard disease controling, and provided support for the research and practice of image recognition methods based on deep learning with small sample sets.

Key words: peach diseases, image recognition, deep learning, DenseNet model

摘要:

为解决传统人工识别桃树病害效率低、成本高、准确率低等问题,提出了基于AI深度学习的桃树病害智能识别方法,利用并微调ImageNet预训练的DenseNet-169分类模型,对桃树常见的11种病害图像进行预处理与模型训练,搭建桃树病害智能识别软件环境。该方法对常见桃树病害的平均识别率达到91%以上,结合图像处理、深度学习、数据挖掘等技术自动对桃树病害进行识别,实现桃树病害的智能诊断并提供防治建议。该方法具有人力成本低、操作简单、识别效率高等优点,利于病害的及时诊出与防治决策的制定,对促进果园病害防控的智慧化管理具有重要研究意义与应用价值。

关键词: 桃树病害, 图像识别, 深度学习, DenseNet模型

CLC Number: