Muhammad Nuzul Naim Baharuddin¹, Hassan Mehmood Khan¹, Norrima Mokhtar¹, Wan Amirul Wan Mahiyiddin¹ Heshalini Rajagopal², Tarmizi Adam³, Jafferi Jamaluddin^4
1Department of Electrical Engineering, Faculty of Engineering, University of Malaya, Malaysia
²Institute of Computer Science and Digital Innovation, UCSI University, 56000 Kuala Lumpur, Malaysia
³Faculty of Engineering, School of Computing, Universiti Teknologi Malaysia 81310 Johor Bahru, Johor, Malaysia
⁴UM Power Energy Dedicated Advanced Centre, Universiti Malaya, 50603, Kuala Lumpur Malaysia
pp. 155–162
ABSTRACT
In recent decades, the rapid growth of urbanization and industrialization has resulted in a significant increase in solid waste, creating an urgent issue that demands attention. The accumulation of solid waste poses a significant challenge, as it can lead to environmental pollution. Recycling is a viable solution that offers economic and environmental benefits. To address this challenge, various intelligent waste management systems and methods are necessary. This research paper explores the use of image processing techniques to classify different types of recyclable dry waste. The study proposes an automated vision-based recognition system that includes image acquisition, feature extraction, and classification. The intelligent waste material classification system extracts 11 features from each dry waste image. The study employed four classifiers - Quadratic Support Vector Machine (Q-SVM), Cubic Support Vector Machine (C-SVM), Fine K-Nearest Neighbor (Fine KNN), and Weighted K-Nearest Neighbor (Weighted KNN) - to categorize the waste into distinct classes, such as bottle, box, crumble, flat, cup, food container, and tin. Among these, the C-SVM classifier performed impressively well, achieving an accuracy of 83.3% and 81.43% during training and testing, respectively. This classifier exhibited consistent performance and had a shorter computation time, making it a highly effective method. Although using the Speeded-Up Robust Features (SURF) method could enhance the classification process, it may lead to longer response and computation times.

ARTICLE INFO
Article History
Received 11 November 2021
Accepted 28 March 2023

Keywords
Support vector machine
Recycling
Feature extraction
Classification

JAALR3306

Download article(PDF)