Research On Grinding Surface Roughness Measurement Methods Based On Machine Vision

Surface roughness refers to the unevenness of a processed surface on which there are minuscule peaks and valleys separated by relatively small spacings. Surface roughness is closely related to the service performance of the workpiece and has a significant impact on the service life and reliability of a mechanical product. Accurate and efficient roughness measurements are of great importance to modern industry. Along with the development and advancement of modern industry and science technology, people put forward higher and higher request on the performance of surface roughness measurement technology, such as high measurement efficiency, high measurement accuracy and good flexibility. Traditional roughness measurement methods have been unable to meet the practical needs. Therefore, to explore lossless, high-efficiency and high-precision roughness measurement methods have become a hot topic of research scholars. Machine vision-based measurement technologies have such advantages as high measurement efficiency, the capacity to acquire large amounts of information, high measurement accuracy, good flexibility, non-contact mode, and high performance-price ratio. Although numerous researchers have used machine vision-based measurement technologies to measure roughness and good results have been achieved, there are still a series of problems need to be solved.Supported by the National Natural Science Foundation of China (Grant No. 71271078), and the Key Project of Science and Technology of Changsha (Grant No. K1306007-11-1), this work studied and explored the grinding surface roughness measuring method based on machine vision.The main research works of this work are as follows:(1) The definition of surface roughness as well as the difference between surface roughness, surface waviness and surface form error were introduced. Based on the national and international standards, the assessment of surface roughness parameters were introduced, and the selection methods of sampling length and evaluation length were described. The general roughness measurement scheme based on machine vision was summarized, and then, the selection skills of light source, lens and cameras were analyzed. The mechanism of imaging of rough surfaces and the effect of different texture directions of a grinding specimen block on the light source image were analyzed.(2) A novel detection method of roughness based on image quality algorithms is proposed to solve the problems that the workpiece measurement range of the current detection methods of surface roughness based on machine vision is limited, and the indices are not comprehensive. A novel image quality index based on regional contrast and gradient similarity (RCGSSIM) is introduced. In RCGSSIM, regional contrast is introduced to GSSIM for more accuracy and calculation method is improved for faster computing. Furthermore, a new grinding surface roughness measurement setup based on image quality is designed. Experimental results show that there is a significant correlation between grinding surface roughness and image quality, and the proposed RCGSSIM has the best comprehensive performance and is more suitable for non-contact online measurement of roughness, when compared with conventional image quality algorithms.(3) A ground surface roughness measurement method is proposed to address current problems in the use of machine vision technology to measure roughness:the calculations are complex, and the measurement process is largely affected by the light source. A reference illumination condition containing two base color light sources for grinding specimens was designed based on the difference in area of the diffusion region between the virtual images formed by a light source on surfaces with different roughness levels. Red, green and blue color space-based color distribution statistical matrices, as well as corresponding overlap degree indexes, are proposed. A model of the relationship between overlap degree index and roughness was constructed. The effect of light source brightness and texture direction on the relationship model is discussed based on the experimental data. The results demonstrate that the surface roughness measurement method, which is based on the overlap degree of the color image, has relatively high accuracy and a relatively wide measurement range and is, to a certain degree, robust to the brightness of the light source and the texture direction. The surface roughness measurement method has huge potential for engineering applications.(4) A novel method, generating reflection images of roughness surface by simulation, is proposed for the time-consuming and labor-consuming of sampling making and the inaccuracy of roughness detecting based on machine vision. Different statistical characteristic cloud data of grinding surface point based on numerical computation are generated, and these data are imported to CAD/CAM to produce 3D entity of grinding surface which then is imported to optical simulation software to produce simulation images under specific experimental condition. Finally, taking the index of color overlap degree as an example, the feasibility of these images which are used to verify the specific index is tested. The experiment results show that reflection images of roughness surface generated by simulation can be used to verify the characteristics of the color overlap degree.