Akademik Veri Yönetim Sistemi
The 5th International Conference on Artificial Intelligence, Smart Technologies and Engineering Applications, Bologna, İtalya, 25 - 27 Nisan 2025, (Yayınlanmadı)
This study presents a novel hybrid waste management system integrating ultrasonic sensor measurements, vision-based inspection, image processing, and artificial intelligence-based classification. The proposed approach addresses the inefficiencies associated with traditional, schedule-based waste collection by accurately determining waste container fill levels, thus optimizing collection routing and timing. A prototype was developed and tested at Yıldız Technical University’s Davutpaşa Campus, demonstrating reduced operational costs, decreased fuel consumption, and a lower environmental impact. Ultrasonic sensors placed on waste containers transmit real-time fill-level data wirelessly via nRF24L01 modules to a Raspberry Pi-based central processing unit, where a regression model estimates container fullness. Additionally, an AI-driven image classification system utilizing a YOLOv8 model cross-validates sensor data, effectively identifying anomalies and mitigating sensor inaccuracies. The system achieved a recognition accuracy of 94%, highlighting its potential for enhancing sustainability in urban waste management.