UAV-based real-time detection of corn earworm using EfficientNet and machine learning.

Area of Science:

• Agricultural Science
• Artificial Intelligence
• Remote Sensing

Background:

• Traditional corn earworm monitoring is labor-intensive and inefficient.
• Early detection is vital for minimizing crop damage and ensuring agricultural productivity.
• Existing methods struggle with timely and accurate pest identification.

Purpose of the Study:

• To develop a real-time, AI-driven system for detecting corn earworm (Helicoverpa zea) infestations using UAVs.
• To leverage multispectral and thermal imaging combined with advanced AI for improved pest detection.
• To provide a scalable solution for precision agriculture and sustainable crop protection.

Main Methods:

• Collected multispectral and thermal images from corn fields throughout the 2024 growing season.
• Utilized image preprocessing techniques including normalization, augmentation, and segmentation.
• Employed EfficientNet for feature extraction and a hybrid Random Forest/Support Vector Machine model for classification.

Main Results:

• Achieved over 90% classification accuracy in detecting corn earworm infestations.
• Demonstrated real-time inference times suitable for practical field application.
• Validated the system's applicability under variable ecological conditions through field trials.

Conclusions:

• UAV-based AI models offer accurate and efficient detection of corn earworm, aiding proactive pest management.
• The methodology is adaptable for other pests and crops, promoting precision agriculture.
• Integrating AI, remote sensing, and entomological validation advances data-driven pest management strategies.