Scholarlink

<p>Field-verifiable technologies for the detection and mapping of diseases in vegetable crops are vital for undertaking precision agriculture practices. The evolving hyperspectral sensors have the capacity to offer plant referenceable spectral data required to map and monitor crop diseases. Tomato is one of the most widely grown vegetable crops in India. Fusarium wilt is a&#xa0;fungal infection that causes severe damage to the growth and yield of tomato crops. As part of the research efforts on developing a&#xa0;regional-level remote sensing system for crop disease surveillance and monitoring, we have undertaken multiple studies pertaining to theoretical modelling, reference spectral data acquisition, and methods for the analyses of hyperspectral data. The objective of this work is the assessment of the spectral discrimination and classification of healthy and Fusarium wilt-infected tomato plants using hyperspectral data. In-situ reflectance spectra of healthy and infected plants over a&#xa0;tomato-growing region (Tumakuru, India) were measured, processed, and analyzed to differentiate between diseased and healthy tomato plants spectrally. We applied nine different methods belonging to machine learning, statistical, and spectral matching approaches considering five different levels of disease severity. Results suggest the existence of stable spectral features which differentiate healthy and diseased plants at distinct levels of disease severity. The prospect of discriminating healthy tomato plants against infected plants with different infection levels is reasonable, as indicated by the different accuracy metrics indicating about 80% accuracy. It is apparent that not all levels of disease severity can be identified spectrally. The detection of disease in plants with invisible symptoms is moderate and is substantially influenced by the method used. With an appropriate combination of methods and disease severity, hyperspectral data-based approaches enable large-scale mapping of Fusarium wilt disease in tomato crop.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hyperspectral Detection and Differentiation of Various Levels of Fusarium Wilt in Tomato Crop Using Machine Learning and Statistical Approaches

  • 期刊论文

  • 发表日期：

    2025年2月5日

查看全文
  • Sivaganesh,
  • Chaitra H.,
  • Rama Rao Nidamanuri,
  • R. G. Sharathchandra,
  • Priya Narayanan

摘要

Field-verifiable technologies for the detection and mapping of diseases in vegetable crops are vital for undertaking precision agriculture practices. The evolving hyperspectral sensors have the capacity to offer plant referenceable spectral data required to map and monitor crop diseases. Tomato is one of the most widely grown vegetable crops in India. Fusarium wilt is a fungal infection that causes severe damage to the growth and yield of tomato crops. As part of the research efforts on developing a regional-level remote sensing system for crop disease surveillance and monitoring, we have undertaken multiple studies pertaining to theoretical modelling, reference spectral data acquisition, and methods for the analyses of hyperspectral data. The objective of this work is the assessment of the spectral discrimination and classification of healthy and Fusarium wilt-infected tomato plants using hyperspectral data. In-situ reflectance spectra of healthy and infected plants over a tomato-growing region (Tumakuru, India) were measured, processed, and analyzed to differentiate between diseased and healthy tomato plants spectrally. We applied nine different methods belonging to machine learning, statistical, and spectral matching approaches considering five different levels of disease severity. Results suggest the existence of stable spectral features which differentiate healthy and diseased plants at distinct levels of disease severity. The prospect of discriminating healthy tomato plants against infected plants with different infection levels is reasonable, as indicated by the different accuracy metrics indicating about 80% accuracy. It is apparent that not all levels of disease severity can be identified spectrally. The detection of disease in plants with invisible symptoms is moderate and is substantially influenced by the method used. With an appropriate combination of methods and disease severity, hyperspectral data-based approaches enable large-scale mapping of Fusarium wilt disease in tomato crop.