AIDM-Strat: Augmented Illegal Dumping Monitoring Strategy through Deep Neural Network-Based Spatial Separation...

Area of Science:

• Computer Science
• Environmental Science
• Artificial Intelligence

Background:

• Increased living standards in South Korea have led to greater waste generation.
• A volume-based trash disposal system was implemented, but citizen participation remains low due to inconvenience, leading to illegal dumping.
• There is a need for automated systems to detect and report illegal waste disposal activities.

Purpose of the Study:

• To develop and evaluate a deep neural network-based system for automatically detecting and reporting illegal garbage dumping.
• To enhance the efficiency and accuracy of waste management monitoring systems.
• To address the challenges posed by passive citizen participation and illegal waste disposal.

Main Methods:

• Utilized OpenPose to extract human articulation points (joints) of individuals disposing of waste.
• Employed the You Only Look Once (YOLO) object detection model to identify garbage bag types.
• Integrated a multi-object tracking (MOT) model to track waste bag IDs and reduce false positives.

Main Results:

• The proposed system accurately determines illegal dumping by analyzing the dumper's joint positions and garbage bag type.
• The multi-object tracking model effectively reduced false detections of illegal dumping.
• Comparative analysis showed the developed system achieved higher accuracy and fewer false alarms than existing behavior recognition systems.

Conclusions:

• The deep neural network-based system offers a highly accurate and reliable solution for monitoring illegal garbage dumping.
• The system's effectiveness in reducing false alarms makes it a valuable tool for environmental protection and waste management.
• This approach has significant potential for various future applications in automated monitoring and enforcement.