Detection of Empoasca Pest Attacks on Tea Leaves Based on Multispectral Camera Using ResNet-50

Tea is one of the important plantation commodities in the Indonesian economy. However, some challenges during cultivation can cause a decrease in production, including the attack of Empoasca pests on tea leaves. Inappropriate use of pesticides to combat this pest can reduce the quality of tea leaves. Accurate diagnosis of plant diseases and proper use of plant protection systems can help prevent production losses. This research proposes to detect Empoasca-infested tea leaves using multispectral cameras, with ResNet-50 architecture used for data processing to achieve high accuracy. The main indicator of Empoasca infestation is leaf curl-up, which can lead to leaf death. Multispectral cameras were chosen for this study because of their ability to provide different spectral features. Among the four spectra tested, the green spectrum gave the best results, graphically showing no overfitting and underfitting. This spectrum makes it easy to identify the characteristics of Empoasca-infested tea leaves based on their shape, which can facilitate data processing using the ResNet-50 architecture.