Optimal method for image detection based on surface roughness of turning parts

In order to realize the purpose of surface roughness detection for turning parts, a new image processing method for surface roughness detection based on machine vision is proposed. Firstly delete part of collected image that severely affected by diffraction according to corresponding algorithm, and then optimize regions according to gray distribution, so as to obtain image grey feature parameters, which can reflect effective feature areas of surface roughness value. Five turning samples with Ra nominal value ranging from 0.8 μm to 12.5 μm are tested using this method. Feature parameters such as mean value, variance, energy and entropy of processed image have a remarkable monotonic relationship with Ra nominal value. The nonlinear error of each feature parameters relationship curves are all within 1.5%. Contrast experiment results show the method can be applied to distinguish and detect surface roughness.