Hyperspectral imaging and deep learning for scrap material characterization

Overview

This innovative system leverages the power of hyperspectral imaging (HSI) combined with deep learning to revolutionize the sorting of scrap materials. By capturing rich spectral information beyond the visible spectrum, the system distinguishes between ferrous and non-ferrous materials, including aluminum, copper, and brass, based on their unique reflectance signatures. This approach promises to surpass traditional sorting methods in accuracy and adaptability, especially under the challenging conditions found in industrial environments.

Technical specifications

• Hyperspectral imaging: Captures detailed spectral data for fine-grained material discrimination.
• Deep learning integration: Utilizes advanced models to identify and classify materials with high precision.
• Real-time processing: Optimized for real-time inference using lightweight architectures and potential hardware acceleration.
• Industrial applicability: Tested under simulated industrial conditions with adjustable parameters for speed, lighting, and dust resistance.
• Adaptability: Capable of adapting to new materials and scrap types, ensuring robust performance across various scrap streams.

Technology readiness level

The system is currently at Technology Readiness Level 4. Initial feasibility has been demonstrated in controlled environments, and the next steps involve real-world validation tests and integration with industrial partners to refine and optimize the technology further.