Optimized Deep Learning Model for Soybean Leaf Classification in Complex Field Environments

• Published in: Korean journal of crop science pp. 68 - 78
• Main Authors: 강지우, 신태환, 고종한
• Format: Journal Article
• Language: English
• Published: 한국작물학회 01.06.2025
• Subjects: 농학
• ISSN: 0252-9777; 2287-8432

Accurate and efficient monitoring of soybean crops is crucial to ensure a stable yield and improved quality. Traditional manual monitoring methods are labor-intensive and time consuming, highlighting the need for automated solutions. Deep learning models, particularly convolutional neural networks (CNNs), represent a promising approach for precise and efficient crop classification. This study introduced an optimized ResNet50 model for classifying soybean varieties using leaf images captured under complex field conditions. Data were collected from experimental fields at Chonnam National University (Gwangju) and the National Institute of Crop Science (Wanju-gun, Jeollabuk-do) between 2021 and 2022, encompassing 4,827 leaf images and 795 canopy images. The dataset was curated under various weather conditions (sunny, cloudy, and rainy), with backgrounds featuring soil, weeds, and other plants to ensure real-world variability. ResNet50 was chosen as the base model because of its robust feature extraction capabilities, and was further optimized through hyperparameter tuning and transfer learning. The final model achieved a classification accuracy of 94.5%, surpassing those of VGG16, Inception-V3, MobileNetV3, and other ResNet variants. These results underscore the effectiveness of CNN-based models in automating soybean variety classification, reducing the reliance on manual assessments, and contributing to advancements in precision agriculture and innovative farming technologies. KCI Citation Count: 0