Scholarlink

<p>Plant health is very important in the field of environmental sciences, botany and agriculture. Both spectral data and images of the plants can be used to extract plant features, which would be helpful in the early detection of plant stress. This paper presents a unique classification approach on generalized plant dataset collected in the real time environment comprising of 18-band spectrometer readings and the RGB images for generalized classification of plants into healthy and unhealthy categories. For the classification of plant, 3 approaches are explored: Classical ML on spectral data (Structured data), CNN on image data and multimodal approach for the combined dataset. The comparison of the results from the experiment clearly demonstrates that the multimodal approach outperforms other approaches because of generalized feature learning because of generalized dataset. The multimodal approach shows an improvement of five percent in classification accuracy compared to the ML model and shows ten percent in classification accuracy compared to the CNN model. Hence, the proposed model offers a real-time solution for agricultural diagnostics.</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

A Fusion-Based Multimodal Approach for Plant Health Classification Using Ground-Based Spectral and Image Data

  • 期刊论文

  • 发表日期：

    2026年6月22日

查看全文
  • Avadh Ladani,
  • Janki Maradia,
  • Jaiprakash Verma,
  • Sumedha Arora

摘要

Plant health is very important in the field of environmental sciences, botany and agriculture. Both spectral data and images of the plants can be used to extract plant features, which would be helpful in the early detection of plant stress. This paper presents a unique classification approach on generalized plant dataset collected in the real time environment comprising of 18-band spectrometer readings and the RGB images for generalized classification of plants into healthy and unhealthy categories. For the classification of plant, 3 approaches are explored: Classical ML on spectral data (Structured data), CNN on image data and multimodal approach for the combined dataset. The comparison of the results from the experiment clearly demonstrates that the multimodal approach outperforms other approaches because of generalized feature learning because of generalized dataset. The multimodal approach shows an improvement of five percent in classification accuracy compared to the ML model and shows ten percent in classification accuracy compared to the CNN model. Hence, the proposed model offers a real-time solution for agricultural diagnostics.