Wheat Pest Identification Based on Deep Learning Techniques

In recent years, effective pest management has become of cross-application of artificial intelligence and big data technologies in agriculture and environment. Nowadays, plant pests have just become a serious problem, which affects the ecological health as well as the growth of crops. In order to further better target traditional crop wheat for pest management, this study utilizes advanced deep learning techniques to accurately detect and categorize common wheat pests (“wireworm”, “Beet spot flies”, and “Cicadellidae”). Through the comparative analysis with the baseline model ResNet-18 model, this study also introduces the more advanced models SE-RegNetY and SEInception-ResNet-V3 models for pest identification in previous studies. Notably, among the three models applied in this study, the SE-Inception-ResNet-V3 model showed excellent results, achieving a significant average overall accuracy of $81.50 \%$ and kappa coefficient of $72.24 \%$, which were $10.3 \%$ and nearly $14.8 \%$ higher than the results of the benchmark ResNet-18 model, respectively. This study marks a further application of deep learning and artificial intelligence in agricultural pest management. By improving the accuracy of agricultural pest identification and classification, it provides the basis and foundation for more sustainable and environmentally friendly pest control methods.