Scholarlink

<p>In agriculture, checking for pests plays an important role, and using machine learning has improved its performance Checking for pests in agriculture is extremely important to preserve crops, control losses, and increase the crop harvest Recognizing pests such as insects, diseases and weeds, on time makes it easier for farmers to select the best methods for handling pests The framework uses various advanced methods to cause it to be accurate and durable Initially, the Wavelet Transform is applied to break the image down into several frequency elements needed for cleaning up the image Thanks to this process, noisy parts are removed while essential details for proper analysis are not lost The next step involves using an Enhanced Deep Convolutional Neural Network (EDCNN) to segment the data The EDCNN has improved layers that help it identify the locations of pests by learning patterns in the images After splitting the image into different pest regions, Hyperspectral Correlation Feature selection is used to get the most essential and distinguishing features It examines the spectrum’s relationships and eliminates any unimportant data to help the model work faster and more efficiently on less data Finally, pictures are classified using a VGG16 model that has been modified and combined with EDCNN The modified version of VGG16 works well on pest images because it exploits the EDCNN’s preliminary segmentation Using the EDCNN on VGG16, we report an accuracy of 99.67%, which is much higher than current technologies Since our suggested approach can be implemented in practice, it promotes further research on identifying pests Kaggle provided us with the proposed method used in our study and a wide variety of images to adequately train and evaluate our results.</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

Advancing Pest Image Detection in Agriculture with Deep Learning and Hyperspectral Correlation Features

  • 期刊论文

  • 发表日期：

    2026年1月7日

查看全文
  • V. Ramya,
  • K. Geetha

摘要

In agriculture, checking for pests plays an important role, and using machine learning has improved its performance Checking for pests in agriculture is extremely important to preserve crops, control losses, and increase the crop harvest Recognizing pests such as insects, diseases and weeds, on time makes it easier for farmers to select the best methods for handling pests The framework uses various advanced methods to cause it to be accurate and durable Initially, the Wavelet Transform is applied to break the image down into several frequency elements needed for cleaning up the image Thanks to this process, noisy parts are removed while essential details for proper analysis are not lost The next step involves using an Enhanced Deep Convolutional Neural Network (EDCNN) to segment the data The EDCNN has improved layers that help it identify the locations of pests by learning patterns in the images After splitting the image into different pest regions, Hyperspectral Correlation Feature selection is used to get the most essential and distinguishing features It examines the spectrum’s relationships and eliminates any unimportant data to help the model work faster and more efficiently on less data Finally, pictures are classified using a VGG16 model that has been modified and combined with EDCNN The modified version of VGG16 works well on pest images because it exploits the EDCNN’s preliminary segmentation Using the EDCNN on VGG16, we report an accuracy of 99.67%, which is much higher than current technologies Since our suggested approach can be implemented in practice, it promotes further research on identifying pests Kaggle provided us with the proposed method used in our study and a wide variety of images to adequately train and evaluate our results.