电子学报 ›› 2019, Vol. 47 ›› Issue (9): 1979-1986.DOI: 10.3969/j.issn.0372-2112.2019.09.023

所属专题: 机器学习之图像处理

• 学术论文 • 上一篇    下一篇

基于高阶残差和参数共享反馈卷积神经网络的农作物病害识别

曾伟辉1,2, 李淼1, 李增2, 熊焰2   

  1. 1. 中国科学院合肥智能机械研究所, 安徽合肥 230031;
    2. 中国科学技术大学, 安徽合肥 230027
  • 收稿日期:2018-08-01 修回日期:2018-09-02 出版日期:2019-09-25
    • 通讯作者:
    • 熊焰
    • 作者简介:
    • 曾伟辉 女,1982年生于陕西咸阳.中国科学院合肥智能机械研究所副研究员,博士研究生.研究方向为计算机视觉与农业信息处理.E-mail:whzeng@iim.ac.cn;李淼 女,1955年生于安徽庐江.中国科学院合肥智能机械研究所研究员,博士生导师.研究方向为人工智能和农业知识工程.E-mail:mli@iim.ac.cn;李增 男,1989年生于河南商丘.中国科学技术大学博士研究生.研究方向为数据挖掘、数据管理和信息安全.E-mail:lizeng@mail.ustc.edu.cn
    • 基金资助:
    • 国家重点研发计划资助 (No.2016YFD0800901-03,No.2017YF0701600)

High-Order Residual and Parameter-Sharing Feedback Convolutional Neural Network for Crop Disease Recognition

ZENG Wei-hui1,2, LI Miao1, LI Zeng2, XIONG Yan2   

  1. 1. Institute of Intelligent Machines, Chinese Academy of Sciences. Hefei, AnHui 230031, China;
    2. University of Science and Technology of China. Hefei, AnHui 230027, China
  • Received:2018-08-01 Revised:2018-09-02 Online:2019-09-25 Published:2019-09-25
    • Supported by:
    • supported by National Key Research and Development Program of China (No.2016YFD0800901-03, No.2017YF0701600)

摘要: 当前,大部分农作物病害图像识别方法主要关注于精度而忽略了鲁棒性.在面向实际环境时,由于噪声干扰和环境因素影响导致识别精度不高.为此提出了一种高阶残差和参数共享反馈的卷积神经网络模型以应用于实际环境农作物病害识别.其中,高阶残差子网络为病害表观提供丰富细致的特征表达,以提高模型识别精度;参数共享反馈子网络用来进一步抑制原深层特征中的背景噪声,以提高模型的鲁棒性.实验结果表明,当面向实际环境农作物病害识别时,本文方法在识别精度和鲁棒性上均优于其他方法.

关键词: 高阶残差, 参数共享反馈, 鲁棒性, 农作物病害识别

Abstract: Most of current crop-disease recognition approaches mainly focus on improving the recognition accuracy on public datasets, while ignoring the recognition robustness.When dealing with real-world recognition problem, the actual recognition accuracy of those approach are often unsatisfactory because of noise interference and environmental influence. To address these issues, we propose a high-order residual and parameter-sharing feedback convolutional neural network (HORPSF) for crop-disease recognition. The high-order residual subnetwork is helpful for improving the recognition accuracy of crop disease. The parameter-sharing feedback subnetwork can effectively depress the background noises and enhance the robustness of the model. Extensive experiment results demonstrate that the proposed HORPSF approach significantly outperforms other competing methods in terms of recognition accuracy and robustness, especially demonstrating superior performance when dealing with the real-world examples of crop-disease recognition.

Key words: high-order residual (HOR), parameter-sharing feedback (PSF), robustness, crop disease recognition

中图分类号: