Integration Framework for Heterogeneous Data Sources in Smart Waste Management: A Case Study on Data Ingestion Pipelines

This thesis presents an integration framework for heterogeneous data sources in smart waste management developed as part of the ReWaste F project. The research aims to enhance recycling efficiency by building a robust data pipeline that integrates and preprocesses data from multiple sensor types, including RGB cameras, Near-Infrared (NIR) sensors, volume flow sensors, and RFID tracking systems. The goal is to support machine learning and deep learning applications for automated waste sorting and material classification, contributing to a more sustainable circular economy. The methodology involved collecting sensor data, developing a data ingestion pipeline, and evaluating its effectiveness using standard machine learning models. Results demonstrate that the pipeline improves data integration and quality, enabling efficient waste sorting and resource recovery. This research supports the creation of smart, digitalized waste management systems in line with European Union initiatives like the Circular Economy Action Plan.