Measurement Of Surface Roughness Based On The CCD Laser Speckle

Surface roughness is an important parameter to describe the surface quality of work-piece and occupies a crucial position in areas like machining. It greatly affects the operability and lifespan of work-piece which mainly shows in the impact on friction, wear, fatigue strength, resistance to corrosion and matching among components etc. The traditional Off-line detection methods contain comparison method, stylus-graphic method, optical interference method, optical scattering method, microscopic method and so on.The statistical property of laser speckle image is closely related with surface roughness of work-piece. We can analyze speckle’s strength distribution, contrast ratio and movement rule in statistical approach. Numerous scholars at home or abroad have done plenty of researches on roughness, but their research content mainly is comparing the change tendencies of varieties speckle image’s characteristic parameters along with the changing of roughness value instead of measuring that value directly. This classified measure method is becoming an important way to gauge surface roughness because of its advantages of non-contact and high sensitivity.This thesis focuses on the flat grinding sample piece on surface roughness. At first, we specifically intercept the speckle images on the sample piece which are collected through speckle acquisition system. Then we use adaptive Wiener filtering and contrast enhancement technology to preprocess the images, filter speckle noise and enhance its contrast character. After that we process the images, extracting the fractal dimension parameters of speckle and analyzing its second-order statistical property by utilizing self-correlation function and fractal algorithm. At last, we curve-fit the sample set constituted by different surface roughness value and its fractal dimension parameters to build a functional relationship between speckle image and surface roughness value. When we detect surface roughness, we can put the parameter extracted from the processes into the function model we getting from the least-squares curve-fitting and inverting to work out the value of surface roughness.