Scholarlink

<p>The application of artificial intelligence (AI) based model like ResNet50 has been examined for crop disease detection through computer vision algorithms, which has potential integration into remote sensing systems. Detecting diseases early is beneficial to decrease crop damage and for optimized usage of pesticides in precision agriculture. We used a&#xa0;dataset of more than 20,000 crop images labelled. Further, we selected 5650 samples for the model training and testing after preprocessing and augmentation. The accuracy, precision, recall, and F1-score were used to evaluate the models. ResNet50 produced the best performance with an accuracy of 91.3%, precision of 90.1%, recall of 92.5%, and F1 score of 91.3%. A&#xa0;comparison analysis showed ResNet50 has a&#xa0;better outcome than VGG16, SVM, and RF. The cross-validation results also showed the model to be robust with a&#xa0;mean accuracy of 91.3% across five folds. Statistical analysis through ANOVA (<i>p</i> = 0.002) and post-hoc t‑tests confirmed the most differences in performance. Transfer learning based deep learning models were effective in capturing complex patterns of the diseases. Yet, field implementation needs further verification under field conditions. This work provides a&#xa0;reproducible and statistically validated benchmark for crop disease detection from images that will boost AI-enabled precision agriculture systems.</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

Computer Vision–Based Crop Disease Detection with Potential Integration into Remote Sensing Systems

  • 期刊论文

  • 发表日期：

    2026年5月5日

查看全文
  • Idowu Olugbenga Adewumi,
  • Babajide Saheed Kosemani,
  • Bukola Olanrewaju Afolabi

摘要

The application of artificial intelligence (AI) based model like ResNet50 has been examined for crop disease detection through computer vision algorithms, which has potential integration into remote sensing systems. Detecting diseases early is beneficial to decrease crop damage and for optimized usage of pesticides in precision agriculture. We used a dataset of more than 20,000 crop images labelled. Further, we selected 5650 samples for the model training and testing after preprocessing and augmentation. The accuracy, precision, recall, and F1-score were used to evaluate the models. ResNet50 produced the best performance with an accuracy of 91.3%, precision of 90.1%, recall of 92.5%, and F1 score of 91.3%. A comparison analysis showed ResNet50 has a better outcome than VGG16, SVM, and RF. The cross-validation results also showed the model to be robust with a mean accuracy of 91.3% across five folds. Statistical analysis through ANOVA (p = 0.002) and post-hoc t‑tests confirmed the most differences in performance. Transfer learning based deep learning models were effective in capturing complex patterns of the diseases. Yet, field implementation needs further verification under field conditions. This work provides a reproducible and statistically validated benchmark for crop disease detection from images that will boost AI-enabled precision agriculture systems.