A new deep learning-based technique for rice pest detection using remote sensing.

Area of Science:

• Agricultural Science
• Computer Science
• Artificial Intelligence

Background:

• Rice production is crucial globally, facing challenges from pests that impact yield and quality.
• Timely pest identification is essential for effective crop protection strategies in rice cultivation.
• Existing methods struggle with accurate and early detection of specific rice pests.

Purpose of the Study:

• To develop a novel deep learning model for detecting two major rice pests: stem borer and Hispa.
• To utilize Unmanned Aerial Vehicle (UAV) technology for real-time pest monitoring in rice fields.
• To improve the accuracy and efficiency of pest detection in agricultural settings.

Main Methods:

• Images captured by a UAV-mounted camera were processed using filtering, labeling, and color thresholding.
• A modified deep learning model, YO-CNN (Yolo-convolution neural network), was developed and implemented.
• Comparative analysis of existing pre-trained models was conducted alongside the proposed YO-CNN approach.

Main Results:

• The proposed YO-CNN model achieved a high accuracy of up to 0.980 in detecting rice pests.
• The model effectively identifies stem borers and Hispa, crucial pests affecting rice crops.
• The study provides a valuable rice pest dataset and demonstrates the superiority of the YO-CNN approach.

Conclusions:

• The YO-CNN model offers a precise and efficient solution for early rice pest detection using UAVs.
• This technology can significantly reduce rice wastage by enabling regular pest monitoring.
• The system supports targeted spraying, optimizing resource use and minimizing environmental impact.