Scholarlink

<p>Insect infestations significantly impede the growth and yield of walnut fruits. Accurate classification remains challenging for farmers due to the similar morphology of early-stage insects and their reliance on visual inspection, which is time-consuming, laborious, subjective, and prone to error. Delayed intervention leads to rapid population proliferation, increases control costs, exacerbates economic risk, promotes synergistic interactions with plant pathogens, and potentially results in yield and productivity loss, which collectively exacerbate the economic burden on farmers. To address the complex insect challenges in walnut fruit, this study proposes an efficient convolutional neural network (CNN) model to classify insect species using a&#xa0;novel walnut fruit insect dataset collected from both experimental and local agricultural fields within and around Handwara, Kupwara district, Jammu and Kashmir, which comprises 17&#xa0;insect classes representing various species. The dataset was pre-processed using both auto and hands-on approaches, ranging from eliminating the background to augmentation operations such as flipping, zooming, translation, scaling, and rotation. InceptionV3, MobileNetV2, VGG16, and ResNet50 models were adapted and fine-tuned for walnut fruit insect classification, incorporating global average pooling and dropout layers to reduce overfitting. Precision, recall, F1 score, receiver operating characteristic (ROC) curves, and confusion matrices were used to evaluate the rigorous performance. Among the four adopted models, the customized InceptionV3 model for walnut fruit insect classification outperformed the others, achieving 97.98% accuracy and 97% precision, recall, and F1 score. Results and findings from experiments demonstrate that the proposed technique is effective for recognizing destructive walnut fruit insects, highlighting its potential to improve insect management and pave the way for sustainable farming practices through mobile applications utilizing smartphone images to detect insects in real time.</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

Application of Different Transfer Learning Models to Classify Destructive Insects in Walnut Fruit

  • 期刊论文

  • 发表日期：

    2025年9月16日

查看全文
  • Khalil Ahmed,
  • Mithilesh Kumar Dubey,
  • Devendra Kumar Pandey

摘要

Insect infestations significantly impede the growth and yield of walnut fruits. Accurate classification remains challenging for farmers due to the similar morphology of early-stage insects and their reliance on visual inspection, which is time-consuming, laborious, subjective, and prone to error. Delayed intervention leads to rapid population proliferation, increases control costs, exacerbates economic risk, promotes synergistic interactions with plant pathogens, and potentially results in yield and productivity loss, which collectively exacerbate the economic burden on farmers. To address the complex insect challenges in walnut fruit, this study proposes an efficient convolutional neural network (CNN) model to classify insect species using a novel walnut fruit insect dataset collected from both experimental and local agricultural fields within and around Handwara, Kupwara district, Jammu and Kashmir, which comprises 17 insect classes representing various species. The dataset was pre-processed using both auto and hands-on approaches, ranging from eliminating the background to augmentation operations such as flipping, zooming, translation, scaling, and rotation. InceptionV3, MobileNetV2, VGG16, and ResNet50 models were adapted and fine-tuned for walnut fruit insect classification, incorporating global average pooling and dropout layers to reduce overfitting. Precision, recall, F1 score, receiver operating characteristic (ROC) curves, and confusion matrices were used to evaluate the rigorous performance. Among the four adopted models, the customized InceptionV3 model for walnut fruit insect classification outperformed the others, achieving 97.98% accuracy and 97% precision, recall, and F1 score. Results and findings from experiments demonstrate that the proposed technique is effective for recognizing destructive walnut fruit insects, highlighting its potential to improve insect management and pave the way for sustainable farming practices through mobile applications utilizing smartphone images to detect insects in real time.