Integration of Consumer Electronics in Precision Agriculture Using Low-Carbon Wheat Lodging Drone Images

Agricultural consumer electronics, such as drones, sensors, and robotics, play a pivotal role in addressing challenges like wheat lodging, which can significantly impact crop yield and quality. This study leverages consumer-grade AAVs to classify wheat lodging types—root lodging and stem lodging—using high-resolution RGB images captured at three altitudes (15, 45, and 91 meters). By employing automatic segmentation techniques, datasets were generated for each altitude, and a refined EfficientNetV2-C model was proposed for classification. The model incorporates a Coordinate Attention (CA) mechanism to enhance feature extraction and Class-Balanced Focal Loss (CB-Focal Loss) to address data imbalance, achieving an average accuracy of 93.58%. This research highlights the integration of advanced AI-based classification with low-carbon agricultural drones, underscoring their relevance to consumer electronics. Compared to four conventional machine learning and two deep learning models, EfficientNetV2-C demonstrated superior performance at all altitudes while maintaining minimal carbon emissions. The study also examines the influence of AAV flight altitude on classification efficacy, revealing that while machine learning models were unaffected, deep learning models showed reduced performance at higher altitudes due to feature loss. These findings emphasize the potential of AAVs as accessible, scalable, and sustainable tools for real-time agricultural monitoring in precision farming.

Hali tarjima qilinmagan