Development of an AI-Enhanced Bubble Curtain System for Automated River Waste Collection and Monitoring

Abstract

Riverine plastic and floating waste continue to threaten aquatic ecosystems and public health in the Philippines, where weak waste management infrastructure and high single-use plastic consumption contribute significantly to river pollution and downstream marine contamination. Existing interception methods such as manual cleanup operations, floating barriers, and mechanical interceptors remain inefficient, labor-intensive, and incapable of real-time monitoring or adaptive response to changing river conditions. This study aimed to design and develop an AI-Enhanced Bubble Curtain System for Automated River Waste Collection and Monitoring that integrates bubble curtain technology with artificial intelligence to automate waste interception, classification, and centralized data management. The system was developed using Agile methodology and the Iterative Design and Development framework, with a Django REST Framework backend, PostgreSQL database, YOLOv8-based image classification model, and a web-based monitoring dashboard built with HTML, JavaScript, and Tailwind CSS. Hardware components were simulated using Wokwi, incorporating an Arduino Uno microcontroller with DHT22, HC-SR04 ultrasonic, water level, and microphone sensors to approximate physical deployment conditions. Results demonstrated that the system successfully performed real-time waste detection and classification through a live camera feed, achieving automated identification of floating debris with confidence scores recorded across 17 captured images, a garbage detection rate of 58.3 percent, and a fully functional administrative dashboard consolidating detection summaries, alert notifications, and user management controls. The study concludes that the proposed system is technically feasible as an intelligent and automated river waste monitoring solution, and recommends future integration of physical hardware components, expanded AI training datasets, and deployment testing in actual riverine environments to validate real-world performance.