Development of online classification system for construction waste based on industrial camera and hyperspectral...

Area of Science:

• Environmental Science
• Materials Science
• Computer Vision

Background:

• Construction waste poses significant environmental and resource management challenges.
• Effective classification is crucial for recovering valuable materials from waste streams.
• Existing methods lack the efficiency needed for real-time waste processing.

Purpose of the Study:

• To develop an efficient online classification system for construction waste.
• To leverage near-infrared hyperspectral imaging for material identification.
• To enhance classification accuracy using advanced machine learning algorithms.

Main Methods:

• An online system integrating an industrial camera and a near-infrared hyperspectral camera was developed.
• Spectral information was captured for objects within a defined region of interest.
• Classification models were built using Extreme Learning Machine (ELM) and Resemblance Discriminant Analysis (RDA).
• An online Particle Swarm Optimization Extreme Learning Machine (PSO-ELM) was implemented for further optimization.

Main Results:

• The integrated hyperspectral and industrial camera system demonstrated efficient classification of construction waste.
• ELM and RDA proved effective for classifying construction waste based on spectral data.
• PSO-ELM further improved the performance and efficiency of the classification system.

Conclusions:

• Near-infrared hyperspectral imaging, combined with industrial cameras, offers an efficient solution for construction waste classification.
• ELM and RDA are viable methods for construction waste material identification.
• PSO optimization enhances the capabilities of ELM for improved waste management systems.