# YOLO-based real-time floating debris counting in urban rivers for flood monitoring and water resource management

## TL;DR

This paper introduces a real-time system using YOLO to count floating debris in urban rivers, helping monitor floods and manage water resources more effectively.

## Contribution

A scalable, real-time debris monitoring system using YOLOv7 for urban rivers, supporting flood risk assessment and SDGs.

## Key findings

- YOLOv7 outperformed YOLOv9 in detecting floating debris with higher precision, recall, and F1-scores.
- The system proved robust under diverse lighting and debris conditions in real-world field tests.
- The framework provides actionable data for flood monitoring and supports SDGs 11 and 13.

## Abstract

Urban flooding is increasingly exacerbated by the accumulation of floating debris in rivers, which obstructs water flow, degrades water quality, and poses significant risks to human safety and environmental sustainability. Effective monitoring of floating debris is therefore critical for early flood warning and long-term water resource management. This study presents a real-time monitoring framework that integrates deep learning-based object detection models, You Only Look Once (YOLO) with video surveillance for the identification and quantification of floating debris in urban rivers. Field deployments were conducted in flood-prone sites in Shah Alam, Malaysia, to evaluate the system under real-world environmental conditions. Results show that YOLOv7 achieved higher accuracy and robustness across diverse debris classes and lighting conditions compared to YOLOv9, with precision, recall, and F1-scores demonstrating strong detection reliability. Beyond technical accuracy, the system provides timely and actionable information for flood risk assessment, river management, and environmental monitoring. By automating debris detection and quantification, this study contributes to Sustainable Development Goals (SDGs) 11 (Sustainable Cities and Communities) and 13 (Climate Action), offering a scalable monitoring solution for flood-prone regions.