An automated solid waste detection using the optimized YOLO model for riverine management

Area of Science:

• Environmental Science and Engineering
• Robotics and Computer Vision

Background:

• Urbanization leads to increasing solid waste pollution in rivers, posing risks to human health and ecosystems.
• Effective riverine management in urban areas necessitates innovative solutions for garbage control.
• Autonomous cleaning robots offer a promising approach to mitigate river pollution.

Purpose of the Study:

• To develop an automated system for detecting floating river garbage crucial for robotic cleaning platforms.
• To enhance the visual system of river cleaning robots through accurate object identification.

Main Methods:

• An improved You Only Look Once (YOLO) object detection model was developed for identifying floating garbage.
• The model was designed to handle challenging conditions including fluctuating illumination, complex backgrounds, and occlusion.
• Focus on strengthening non-linear feature extraction to improve detection accuracy.

Main Results:

• The proposed YOLO model demonstrated rapid convergence, reducing training time.
• The system achieved a mean average precision (mAP) of 89% for garbage detection.
• Successfully identified five distinct classes of riverine waste: plastic bottles, aluminum cans, plastic bags, styrofoam, and plastic containers.

Conclusions:

• The developed automated detection system is feasible for real-time identification of floating river garbage.
• This technology is vital for the development of efficient robotic platforms for river cleaning.
• The improved YOLO model offers a robust solution for tackling urban river pollution.