Inspection of complex surfaces by means of structured light patterns

Area of Science:

• Computer Vision
• Metrology
• Surface Engineering

Background:

• Industrial surface inspection often uses structured light for characterization.
• Existing methods are primarily suited for simple geometries like cylindrical surfaces.
• Generalizing these techniques to complex free-form shapes is a significant challenge.

Purpose of the Study:

• To generalize a structured light-based surface inspection methodology.
• To adapt the technique for characterizing free-form rough and specular surfaces.
• To develop a general approach for free-form stripe image interpretation.

Main Methods:

• A four-step procedure for free-form stripe image interpretation was proposed.
• Techniques for feature-based image content description were compared.
• Optimal feature sub-groups were determined, fused, and selected.
• Pattern recognition processes were defined and optimized.

Main Results:

• The generalized methodology successfully characterizes free-form rough and specular surfaces.
• Feature selection increased classification rates by over 2% compared to the initial fused set.
• Similar high classification rates were achieved using approximately 25% of the initial features.

Conclusions:

• The proposed stripe image interpretation approach is effective for free-form surfaces.
• Optimized feature selection enhances inspection accuracy and efficiency.
• This generalization expands the applicability of structured light inspection techniques.