Artificial Intelligence and Hyperspectral Imaging in Shoe Recycling

New Publication by Avicon and VIVE Textile Recycling Experts

We’re pleased to announce that the latest issue of Pomiary Automatyka Robotyka (PAR) features an article authored by experts from Avicon and VIVE Textile Recycling. The publication focuses on leveraging deep learning and hyperspectral imaging technologies in the processes of shoe recycling.

Challenges in Shoe Recycling – Why Is Material Classification So Difficult?

Rising environmental awareness and the need to reduce waste pose new challenges for the recycling industry. One of the critical issues is the accurate identification of materials used in shoe components.

Traditional vision systems based on mono or RGB cameras are insufficient to distinguish materials such as leather, rubber, plastics, or textiles effectively. The challenge becomes even more complex with second-hand footwear, which presents significant variability in colors, textures, and wear levels.

Hyperspectral Imaging and Deep Learning – A New Standard in Material Classification

In response to these challenges, our team developed a system combining hyperspectral imaging (HSI) technology in the NIR-SWIR range (900–1700 nm) with deep learning methods. This combination enables significantly higher accuracy in material classification than conventional vision technologies.

The Avicon Hyperspectral System – Solution Details

Measurement Range and NIR-SWIR Specifics

The system utilizes Specim FX17 hyperspectral cameras that capture images in narrow spectral bands, enabling the identification of material-specific spectral features invisible to the human eye or standard RGB cameras.

Preparing Training Data

For the project, we created databases of hyperspectral images, which were manually masked to isolate areas corresponding to specific materials. This process required significant precision and expertise from the operators involved.

Test Results and Solution Effectiveness

Tests confirmed high accuracy in classifying shoe materials, achieving an average accuracy above 80% on the test set. Particularly noteworthy is the consistency of results, crucial in industrial conditions.

Benefits of Using AI in Shoe Sorting

Implementing hyperspectral and AI technologies provides:

• significant reduction in sorting costs,
• increased precision in material classification,
• reduced waste sent to landfill or incineration,
• support for circular economy initiatives.

Learn More

We invite you to read the full article and contact our team to discuss how hyperspectral solutions could be integrated into your production or recycling processes.

Download the article: https://doi.org/10.14313/par_256/85