Vision Based Detection of Mealybug Infection in Custard Apple Using Machine Learning

Abstract

Insects and pests have always been a major factor hampering the crop outcome and degrading its market price. The custard apple is one such fruit that faces the wrath of pests reducing its market value and net yield. Early detection of such pests can largely help farmers to plan a proper mechanism to fight them and reduce the damage or the orchard. This paper primarily focuses on the early detection of mealybugs in custard apples. A mealybug is a cotton-like bug that inhibits the exterior of the fruit and feeds on the fruit sap. This not only reduces the nutrition of the fruit but also makes fruit look unpleasant thus reducing its market price. Many models including CNN, Random Forest, Xgboost, and SVM have been implemented on the collected dataset. The system developed is able to differentiate mealybugs on green custard apples and classify them as infected or uninfected. Models like SVM, Random Forest, KNN, and Xgboost have been implemented on the dataset.