Planococcus citri in coffee trees by supervised classification using multispectral images

ABSTRACT: One of the obstacles to the cultivation of coffee trees is the occurrence of the mealybug Planococcus citri. These insects occur in clusters, and can infest branches, leaves, flower buds, and fruits, reaching all the rosettes of the plant and leading to partial drying or total loss of these branches. This study determined the use of low-cost multispectral images to identify coffee plants infested with P. citri. Two study areas were used in the municipality of Coromandel, Minas Gerais, Brazil. One had a high infestation rate of mealybugs and the other had no mealybugs. Another area, in the municipality of Monte Carmelo, Minas Gerais, Brazil, had mealybugs on some plants and not on others. In each study area, 50 plants were randomly sampled, with a minimum distance of 10 m between them. Then, we counted the number of mealybugs in two plagiotropic branches in the middle third portion of the plants. The images were obtained using a drone attached to a Mapir Survey 3W camera at a height of 100 m. The classifications were made using Artificial Neural Networks (ANN), Support Vector Machine (SVM), and Random Forest algorithms. The results confirmed the possibility of distinguishing between healthy and P. citri infested plants using algorithms based on machine learning. The best differentiation of healthy and infested plants was achieved using the Random Forest algorithm in areas with variable infestation (EG = 90% and K = 0.80), followed by SVM (EG = 83.34% and K = 0.67), and ANN (EG = 73.34% and K = 0.47).