Surface texture evaluation using 3D reconstruction from images by parametric anisotropic BRDF

Abstract

We present a method for surface texture evaluation using machine vision by studying the phenomenon of reflection from a real surface. A parameterized anisotropic bi-directional reflectance distribution function (BRDF) is proposed along with a fusion reconstruction method which is analogous to the human visual system. The proposed reflection model operates on an image dataset to output the reconstructed shape. Machined surfaces obtained by performing mechanical grinding at varying machining conditions are analyzed using gray level co-occurrence matrix (GLCM), wavelet decomposition, photometric stereo and fusion reconstruction based texture analysis to study and benchmark performance of both statistical and topographical surface texture evaluation methods. The four methods are implemented in MATLAB™ to estimate surface roughness parameters – Ra, Rq, and Rz. Error analysis is performed by comparing estimated roughness values against stylus profilometer measurements. The comparison reveals that fusion reconstruction estimates surface roughness closer to stylus profilometer measurements as compared to GLCM, wavelet decomposition and photometric stereo based texture analysis.