Development and evaluation of a color-image-based visual roughness measurement method with illumination robustness

Area of Science:

• Materials Science
• Computer Vision
• Metrology

Background:

• Industrial surface roughness measurement is often hindered by variations in lighting conditions.
• Existing machine vision methods lack adaptability to illumination changes, limiting their practical application.

Purpose of the Study:

• To develop a machine vision method for roughness measurement that is robust to illumination variations.
• To investigate the role of color image indices in enhancing mathematical models for roughness assessment.
• To evaluate the effectiveness of the proposed method in diverse industrial scenarios.

Main Methods:

• A novel machine vision approach utilizing color image indices was developed.
• Analysis of virtual images of surfaces with varying roughness and texture orientations.
• Derivation of the singular value ratio as a key roughness evaluation index.
• Experimental validation using a support vector machine (SVM) model.

Main Results:

• The proposed method demonstrated consistent index values despite significant illumination changes for both vertical and horizontal surface textures.
• The singular value ratio proved effective in evaluating surface roughness.
• Verification against an SVM-based method confirmed high measurement accuracy (90%) across a range of 0.127-2.245 µm.

Conclusions:

• The developed machine vision method exhibits strong robustness to illumination variations, addressing a key limitation in industrial roughness measurement.
• The method is accurate, feasible, and shows significant potential for practical implementation in engineering contexts.
• The use of color image indices and the singular value ratio offers a promising advancement in non-contact surface metrology.