Melamine Faced Panels Defect Classification beyond the Visible Spectrum.

Area of Science:

• Materials Science
• Computer Vision
• Industrial Quality Control

Background:

• Melamine faced panels are susceptible to various defects during production.
• Accurate defect detection is crucial for maintaining product quality and reducing waste.
• Current defect detection methods often rely solely on visible spectrum imaging, which may miss subtle imperfections.

Purpose of the Study:

• To investigate the effectiveness of multi-spectral imaging for classifying defects in melamine faced panels.
• To compare the performance of visible (VS), near-infrared (NIR), and long wavelength infrared (LWIR) imaging for defect detection.
• To evaluate feature descriptor learning approaches combined with machine learning classifiers for this task.

Main Methods:

• Utilized images from VS, NIR, and LWIR spectral bands.
• Employed feature descriptor learning with Extended Local Binary Patterns (E-LBP) and SURF using Bag of Words (BoW) representation.
• Trained and evaluated a Support Vector Machine (SVM) classifier on a dataset of five industrial defect categories.

Main Results:

• Multi-spectral imaging, incorporating NIR and LWIR, demonstrated superior defect classification performance compared to VS imaging alone.
• The combination of spectral bands provided richer information for distinguishing between different defect types.
• Feature descriptor learning effectively captured relevant textural and structural information for classification.

Conclusions:

• Integrating non-visible spectral bands (NIR, LWIR) enhances defect classification accuracy in melamine faced panels.
• Multi-spectral imaging offers a more robust solution for industrial quality control of these materials.
• The evaluated feature learning and classification methods are effective for automated defect detection.