Investigation of Deep Learning Approaches for Identification of Important Wheat Pests in Central Anatolia

Abstract

Artificial intelligence-based systems play a crucial role in Integrated Pest Management studies. It is important to develop and support such systems for controlling wheat pests, which cause significant losses in wheat production which is strategic importance, particularly in Turkey. This study employed various pre-trained deep learning approaches to identify key wheat pests in the Central Anatolia Region, namely Aelia spp., Anisoplia spp., Eurygaster spp., Pachytychius hordei, and Zabrus spp. The models' classification success was determined using open and original datasets. Among the models, the ResNet-18 model outperformed others, achieving a classification success rate of 99%. Furthermore, each model was tested with original images collected during field studies to assess their effectiveness. The results demonstrate that pre-trained deep learning models can be utilized for the identification of important wheat pests in Central Anatolia as part of Integrated Pest Management.

Keywords

• important wheat pests
• insect classification
• deep learning
• cnn
• transfer learning
• resnet-18