Deep Learning-Based Monitoring of Illegal Plastic Waste Disposal in Urban Areas

Abstract Illegal plastic waste disposal poses a significant environmental and public health challenge in rapidly urbanizing regions. Traditional waste management systems often lack real-time monitoring capabilities, making enforcement and mitigation difficult. This study proposes a deep learning-based surveillance framework that utilizes closed-circuit television (CCTV) footage to automatically detect and classify incidents of plastic waste dumping in urban environments. Leveraging convolutional neural networks (CNNs) and object detection models such as YOLOv8, the system is trained on a custom dataset comprising various forms of plastic waste in diverse urban scenarios. Temporal sequence analysis is incorporated to distinguish between transient pedestrian behaviour and deliberate dumping actions. Experimental evaluations demonstrate high detection accuracy and robustness across varying lighting and weather conditions. The proposed approach offers a scalable, non-intrusive solution for smart city waste monitoring, enabling municipal authorities to respond more efficiently to environmental violations. Future work will explore integration with geospatial alert systems and multi-camera tracking for broader deployment. Keyword: Smart surveillance, YOLO object detection, Computer vision, Convolutional Neural Networks (CNNs)