Detection of Tomato Leaf Miner Using Deep Neural Network.

Area of Science:

• Agricultural Science
• Plant Pathology
• Computer Vision

Background:

• Climate change and global warming are increasing the dispersal rate of plant diseases and pests.
• The tomato leaf miner (Phyllocnistis citrella) causes severe damage, leading to 80-100% crop loss.
• This pest is globally distributed, necessitating effective detection and management strategies.

Purpose of the Study:

• To investigate deep neural network (DNN) approaches for improved tomato leaf miner detection.
• To compare the performance of DNN models for image classification and image segmentation in identifying the pest.
• To evaluate the efficacy of DNN models in real-world agricultural settings.

Main Methods:

• Two DNN models, one for classification and one for segmentation, were trained using RGB images of tomato leaves.
• Images were captured from real-world agricultural sites to ensure practical relevance.
• Model performance was evaluated using precision, recall, and F1-score metrics.

Main Results:

• The DNN model utilizing segmentation demonstrated superior performance compared to the classification model.
• Segmentation model achieved higher precision, recall, and F1-score values for tomato leaf miner detection.
• Crucially, the segmentation model produced zero false negatives, indicating high reliability.

Conclusions:

• Deep neural network-based image segmentation is a highly effective method for detecting the tomato leaf miner.
• The segmentation approach offers a reliable tool for plant disease and pest detection in agriculture.
• This technology can aid in timely interventions to protect tomato crops from significant losses.