Leveraging Machine Learning and Deep Learning Techniques for Accurate Classification of Stored Grain Pests

The present study proposes a methodology utilizing machine learning and deep learning techniques for stored grain insect pest classification. Relevant morphological features extracted from captured pest images were fed to K-nearest neighbors (KNN), Support Vector Machine (SVM), Convolutional Neural Network (CNN), and Naïve Bayes (NB) algorithms. The effectiveness of the proposed approach was evaluated using a comprehensive dataset compiled with images of various stored grain insect pests. The order of classification accuracy was NB < KNN < SVM < CNN where KNN achieved 76% accuracy, SVM exhibiting 81% accuracy, CNN achieving 98% accuracy, and NB achieving 33% accuracy. Though CNN required more computation time for classification, better accuracy was achieved and this could be utilized to identify the insects infesting stored food grains. The intelligent classification provides a valuable tool for identifying and differentiating stored grain insect pests, the primary step in IPM.