A multivariate method for surface roughness vision inspection in different ambient light

Abstract

Machine vision method was used to assess surfaces roughness for different Φ38mm grinding axes in different ambient light conditions. The effect of ambient light was first analyzed on the method based on the gray-level co-occurrence matrix. Then a new multivariate-based method was proposed to minimize the effect of ambient light on inspection. The features of the background and the axis were extracted for corresponding real surface roughness in system calibration. Then the corresponding features of inspected axes were input to calculate the inspecting values. The experimental results show that the inspecting accurate rate of proposed method is improved than the gray-level co-occurrence matrix method when the error between inspecting value and its corresponding real surface roughness is set 0.05μm.