Scholarlink

<p>Agricultural productivity relies heavily on effective crop management, which necessitates accurate differentiation between crop and weed plants. Misclassification in this process can lead to significant economic losses due to inadequate weed control or accidental removal of crops. Methods: This study proposes an ensemble-based classification approach using Convolutional Neural Networks (CNNs) to enhance the accuracy of crop and weed identification. The ensemble incorporates three pre-trained deep learning (DL) models such as Inception-ResNetv2, VGG19, and DenseNet-169 trained on extensive image datasets. A&#xa0;fusion approach based on fuzzy ranks combines decision scores from these models, integrating confidence levels to improve prediction robustness. The model’s performance is evaluated on benchmark datasets, including Weed Detection in Soybean Crops, using a&#xa0;rigorous 5‑fold cross-validation strategy. Additionally, Cosine Similarity (C.S) is applied to analyze and measure the similarity between crops and weeds, providing further insights into plant classification. Results: The specified ensemble approach performs better than the others in weed classification, achieving 99.39% classification accuracy and 99.58% sensitivity in a&#xa0;5-class configuration. These results outperform existing methods for weed classification in soybean crops, highlighting the efficiency of the model in differentiating between crop and weed plants with high precision. Conclusion: The findings of this research emphasize the potential of ensemble-based CNN models in advancing automated weed classification for precision agriculture. By improving classification accuracy and sensitivity, the proposed approach contributes to optimized crop management, increased yield efficiency, and reduced losses caused by invasive weeds. Additionally, the application of Cosine Similarity further enriches the knowledge of plant classification, paving the way for more refined and intelligent agricultural solutions.</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

Weed Detection Approach for Soybeans Using Fuzzy Ensemble and Image Similarity

  • 期刊论文

  • 发表日期：

    2026年1月15日

查看全文
  • G. V. S. Narayana,
  • Sanjay K. Kuanar,
  • Punyaban Patel,
  • S. Gopal Krishna Patro,
  • Ayodeji Olalekan Salau

摘要

Agricultural productivity relies heavily on effective crop management, which necessitates accurate differentiation between crop and weed plants. Misclassification in this process can lead to significant economic losses due to inadequate weed control or accidental removal of crops. Methods: This study proposes an ensemble-based classification approach using Convolutional Neural Networks (CNNs) to enhance the accuracy of crop and weed identification. The ensemble incorporates three pre-trained deep learning (DL) models such as Inception-ResNetv2, VGG19, and DenseNet-169 trained on extensive image datasets. A fusion approach based on fuzzy ranks combines decision scores from these models, integrating confidence levels to improve prediction robustness. The model’s performance is evaluated on benchmark datasets, including Weed Detection in Soybean Crops, using a rigorous 5‑fold cross-validation strategy. Additionally, Cosine Similarity (C.S) is applied to analyze and measure the similarity between crops and weeds, providing further insights into plant classification. Results: The specified ensemble approach performs better than the others in weed classification, achieving 99.39% classification accuracy and 99.58% sensitivity in a 5-class configuration. These results outperform existing methods for weed classification in soybean crops, highlighting the efficiency of the model in differentiating between crop and weed plants with high precision. Conclusion: The findings of this research emphasize the potential of ensemble-based CNN models in advancing automated weed classification for precision agriculture. By improving classification accuracy and sensitivity, the proposed approach contributes to optimized crop management, increased yield efficiency, and reduced losses caused by invasive weeds. Additionally, the application of Cosine Similarity further enriches the knowledge of plant classification, paving the way for more refined and intelligent agricultural solutions.