Abstract

Reduction of agricultural, production is a serious issue in the agricultural sector, largely because of insect attacks on field’s plants. Identifying & categorization of insects have historically been labour-intensive processes that have necessitated the services of trained entomologists. Earlier warning of an insect assault aids farmers in mitigating crop injury, which in turn increases crop yield and decreases pesticide use. To use a variety of features extracted such as texture, colour, form, histogram of oriented gradients-HOG, & global image descriptor, this study classifies crop insects through the application of machine vision & knowledge-based methodologies with image processing (GIST). Insects were organised according to a system that took into account all of these characteristics. In this study, 3 separate insect datasets was subjected to a variety of machine learning-ML methods, such as basic classifiers as well as ensemble classifications, with the results of these classifications being ranked according to a majority vote. Several different types of base classifiers was utilised, including naive bayes-NB, support vector machine-SVM, K-nearest-neighbour-KNN, & multi-layer perceptron-MLP. In order to improve the classification & identifying of insects, they used a combination of ensemble classifications, including random forest-RF, bagging, & XGBoost, as well as we ran a 10-fold cross-validation test. Empirical outcomes demonstrated that using majority voting with ensembles classifications to include texture, colour, shape, HOG, & GIST characteristics enhanced classifications performance.

Description