Design and development of smart Internet of Things-based solid waste management system using computer vision

Area of Science:

• Environmental Science
• Computer Science
• Artificial Intelligence

Background:

• Municipal solid waste (MSW) management is critical for environmental protection, economic sustainability, and public health.
• Current MSW systems often improperly dispose of waste, with approximately 90% ending up in open dumps or landfills.
• Accurate waste classification is essential for effective waste-to-energy conversion and proper disposal.

Purpose of the Study:

• To develop an effective computer vision-based solution for MSW classification using IoT and machine learning.
• To improve the accuracy and reduce the error rate of waste classification compared to existing methods.
• To enable precise segregation of waste into biodegradable and non-biodegradable categories for optimized management.

Main Methods:

• Utilized Internet of Things (IoT) and machine learning (ML) techniques, including regression, classification, and clustering.
• Developed and trained a Convolutional Neural Network (CNN) model, incorporating ResNet V2 architecture and transfer learning.
• Implemented batch normalization and mixed hybrid pooling techniques to enhance CNN stability and performance.

Main Results:

• The proposed ResNet-based CNN model achieved superior performance in waste image classification.
• Demonstrated a 19.08% higher accuracy in waste classification compared to existing models.
• Achieved a 34.97% lower loss rate, indicating improved model efficiency and reliability.

Conclusions:

• The developed computer vision system effectively classifies MSW into biodegradable and non-biodegradable categories.
• The ResNet-based CNN model offers a significant advancement over existing methods for waste management.
• This technology supports precise waste collection, enhancing recycling and waste-to-energy initiatives.