A robust classification algorithm for separation of construction waste using NIR hyperspectral system

Area of Science:

• Materials Science
• Spectroscopy
• Environmental Engineering

Background:

• Construction waste management faces challenges in utilization, cost-efficiency, and processing speed.
• Near-infrared hyperspectral technology offers potential for automated waste identification.
• Existing methods may struggle with data redundancy and distinguishing similar spectral signatures.

Purpose of the Study:

• To develop an efficient method for classifying construction waste using near-infrared hyperspectral technology.
• To improve the accuracy and robustness of construction waste identification under complex conditions.
• To reduce processing costs and enhance the utilization rate of construction waste.

Main Methods:

• Proposed Pythagorean Wavelet Transform (PWT) for characteristic reflectivity extraction, reducing hyperspectral data redundancy.
• Extracted and evaluated additional features: first derivative and intrinsic mode function (IMF).
• Employed Random Forest (RF) for trend-feature identification and Extreme Learning Machine (ELM) for amplitude-feature identification.
• Developed a complementary troubleshooting (CT) method combining ELM and RF for online identification.

Main Results:

• PWT retained more detailed information and enhanced spectral differences compared to standard Wavelet Transform (WT).
• First derivative and IMF were identified as effective features for distinguishing similar spectra.
• The CT method, using ELM for initial identification and RF for verification, improved model robustness.
• Achieved 100% accuracy in classifying 180 samples across six construction waste types (woods, plastics, bricks, concretes, rubbers, black bricks).

Conclusions:

• The proposed PWT-based hyperspectral approach combined with a complementary troubleshooting (CT) method offers a highly accurate and robust solution for construction waste classification.
• This technology can significantly improve construction waste management by enhancing identification accuracy and efficiency.
• The method demonstrates potential for real-world application in automated waste sorting and recycling processes.