Vision-Based On-Site Construction Waste Localization Using Unmanned Aerial Vehicle

Area of Science:

• Environmental Science
• Robotics
• Computer Vision

Background:

• Construction and demolition waste (CDW) significantly contributes to environmental pollution.
• Existing automated recycling methods using robots are limited by speed and detection range.
• Current construction waste recognition techniques are ineffective for Unmanned Aerial Vehicles (UAVs), lacking accurate coordinate conversion.

Purpose of the Study:

• To develop an efficient method for autonomous construction waste localization using UAVs.
• To improve the efficiency of construction waste management and recycling.
• To address the limitations of existing technologies in CDW recognition and localization.

Main Methods:

• Development of a high-precision, long-range construction waste recognition algorithm for UAV imagery.
• Creation of a dedicated image dataset for construction waste.
• Implementation of a novel method to convert pixel coordinates from images to real-world coordinates.

Main Results:

• The proposed method significantly enhances the discernibility of computer vision algorithms for small targets and high-frequency details.
• Achieved approximately 2% improvement in accuracy and recall for construction waste localization using high-resolution UAV imagery.
• Demonstrated effective operation at speeds up to 28 frames per second (fps).

Conclusions:

• The developed UAV-based system offers an efficient solution for construction waste management.
• The method enables accurate localization of construction waste, facilitating improved recycling processes.
• This technology supports the broader green development of the construction industry through enhanced waste management.