Quantitative Evaluation of Surface Roughness of Flat Metal Surfaces Using Computer Vision System

In this paper, a method of quantitative evaluation of surface roughness based on computer vision system is presented. A low cost computer vision system consisting of flat bed desktop scanner connected to personal computer (PC) is used. A large number of surface specimens such as EN-8, EN-9, cast iron, copper, brass, aluminium, C-20, C-45 steel etc. were carefully prepared by using various machining processes like planing, shaping, turning, milling, grinding, polishing etc. to generate a database of surface specimens with different lay-types and surface roughness values. This database is evaluated for conventional surface roughness parameters like R_t, R_a, R_q and for RGB colour component values at each pixel over the digital images of these produced surfaces. By using the technique of multiple linear regression analysis, the conventional roughness values and colur component values were correlated with each other to form a multiple linear regression equation for R_t. The value of surface roughness R_t obtained for a given specimen using this equation was then crosschecked and confirmed with the results obtained by using conventional method for the same specimen. When any test surface is introduced for surface roughness evaluation, the developed method relates the colour component values obtained from its surface image, to the conventional values like R_t, R_a, R_q. In addition to this, surface topographical representation and summits are also presented. Using this method even the evaluation of the surface roughness in the nano-metre level can be carried out to fulfill the requirements of experimental field of 0.001 to 50 microns.