Image Enhancement with DBPN for Plant Pest Classification

In an effort to control plant pests that adversely affect agricultural productivity, accurate early classification and recognition of plant pests are crucial. This study aims to enhance image quality using the Deep Back-Projection Network (DBPN) technique for the classification of plant pests using Convolutional Neural Network (CNN). The dataset consists of 3,150 images of pests from nine different species. The classification process is carried out using a CNN model with four convolutional layers combined with hyperparameter tuning, yielding an accuracy of 94%. Evaluation results indicate that the model has good precision, recall, and F1-score values for most pest classes, although the recall values for mites and sawflies are slightly lower. Enhancing image quality using DBPN results in a 10% increase in classification accuracy. This study demonstrates that better image quality can help the model recognize important features more accurately.