Image classification of sugarcane aphid density using deep convolutional neural networks

• DOI: 10.1016/j.atech.2022.100089
• Publication Date: 2022-06-28T00:40:26Z
• Authors: Ivan Grijalva; Brian J. Spiesman; Brian McCornack
• Keywords: Sweet sorghum; Sprayer

Sugarcane aphid, Melanaphis sacchari (Zehntner), has caused significant yield loss across the sorghum (Sorghum bicolor L. Moench) production region in U.S. Adequate management of sugarcane aphid depends on pest monitoring and economic threshold levels to spray insecticides. However, scouting this under field conditions is time-consuming inefficient. To assist monitoring, we propose use deep learning models automatically classify infestation leaves according different density images. We used a total 5,048 images collected during events evaluated performance four classification models: Inception v3, DenseNet 121, Resnet 50, Xception. trained densities into 6 classes based established standard for spraying: no aphids present (0 aphids/leaf), threat or below an action/treatment (1–10, 11–39 infested above where insecticide should be applied (40–125, 126–500, > 500 aphids/leaf) manage conditions. Among these models, v3 Xception performed best with overall accuracy score 86% lower number misclassified Importantly, correctly classified as over 97% time. The methodology developed tested study can sampling protocols further mobile applications remote sensing technologies. These technologies growers researchers scout screen susceptible resistant varieties provide accurate recommendations whether not apply pesticides.