Scholarlink

This study aims to monitor the growth of major field crops rice, onion, and garlic using drone-based imagery and to develop a CNN-based yield prediction model. High resolution RGB and multispectral images were collected via UAVs, followed by preprocessing steps such as orthorectification and overlap validation to build the training dataset. Semantic segmentation models including U-Net, FCN, and DeepLab v3+ were applied to extract vegetation areas, and their performance was compared. A Multi-input CNN model was then implemented to predict crop yield by simultaneously utilizing RGB and multispectral data. The analysis revealed that DeepLab v3+ outperformed other models in segmentation accuracy, and the use of vegetation index (e.g., NDVI) derived from multispectral imagery significantly improved prediction precision. This research demonstrates the feasibility of accurate growth monitoring and yield forecasting in open-field environments and presents a foundational technology for advancing smart farming and data-driven agricultural decision-making.

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

Development of Multi-input Model for Forecasting Field Crop Yields

  • 章节

  • 发表日期：

    2026年

查看全文
  • Kyeong Il Ko,
  • Hyun Yoe,
  • Meong Hun Lee

摘要

This study aims to monitor the growth of major field crops rice, onion, and garlic using drone-based imagery and to develop a CNN-based yield prediction model. High resolution RGB and multispectral images were collected via UAVs, followed by preprocessing steps such as orthorectification and overlap validation to build the training dataset. Semantic segmentation models including U-Net, FCN, and DeepLab v3+ were applied to extract vegetation areas, and their performance was compared. A Multi-input CNN model was then implemented to predict crop yield by simultaneously utilizing RGB and multispectral data. The analysis revealed that DeepLab v3+ outperformed other models in segmentation accuracy, and the use of vegetation index (e.g., NDVI) derived from multispectral imagery significantly improved prediction precision. This research demonstrates the feasibility of accurate growth monitoring and yield forecasting in open-field environments and presents a foundational technology for advancing smart farming and data-driven agricultural decision-making.