欢迎访问《中国农学通报》,

中国农学通报 ›› 2024, Vol. 40 ›› Issue (4): 158-164.doi: 10.11924/j.issn.1000-6850.casb2022-1025

• 农业信息·科技教育 • 上一篇    

基于多特征的BP神经网络多种植物叶片病害识别研究

马娜(), 任宇翔   

  1. 山西农业大学信息科学与工程学院,山西太谷 030801
  • 收稿日期:2022-12-13 修回日期:2023-08-10 出版日期:2024-02-05 发布日期:2024-01-29
  • 作者简介:

    马娜,女,1992年出生,山西临汾人,讲师,硕士,主要研究方向农业信息化与图像处理研究。E-mail:

  • 基金资助:
    山西省基础研究计划“基于深度学习的育肥猪异常行为检测及个体跟踪方法研究”(202103021223141); 山西农业大学青年科技创新基金“基于深度学习的小麦病害识别与应用研究”(2020QC17)

Identification of Various Plant Leaf Diseases Based on Multi-feature BP Neural Network

MA Na(), REN Yuxiang   

  1. College of Information Science and Engineering, Shanxi Agricultural University, Taigu, Shanxi 030801
  • Received:2022-12-13 Revised:2023-08-10 Published-:2024-02-05 Online:2024-01-29

摘要:

为准确识别植物的健康状况,更好地对植物进行健康管理和治疗,以芒果、柠檬和石榴3种植物健康和病害叶片为研究对象,设计BP神经网络模型对植物健康状况进行识别。首先提取植物叶片表型特征数据,包括叶片颜色特征、形状特征和纹理特征。其中使用小波变换提取植物叶片的纹理特征,并用PCA主成分分析法对提取的特征数据降维。其次建立BP神经网络模型对植物进行分类识别。采用不同特征组合进行实验,识别准确率最高可达83.9%。采用颜色、形状和纹理组合特征建立的BP神经网络植物叶片健康识别模型具有最好的识别效果,可以便捷、高效地识别植物病害。

关键词: 植物叶片, 病害识别, 特征提取, 主成分分析, BP神经网络

Abstract:

In order to accurately identify the health status of plants and better manage the health of plants, this study took healthy and diseased leaves of pomegranate, lemon and mango plants as the research objects, and designed BP neural network model to identify the health status of plants. Firstly, the color and shape features of plant leaves were extracted, then the texture features of plant leaves were extracted by wavelet transform, and the dimension of extracted feature data was reduced by PCA method. Secondly, BP neural network model was established to classify and recognize plants. The recognition accuracy of different feature combinations was up to 83.9%. The BP neural network based on color, shape and texture features has the best recognition effect, and can identify the disease of various plant leaf conveniently and efficiently.

Key words: leaves of plants, disease identification, feature extraction, principal component analysis (PCA), BP neural network