Research On Surface Roughness Measurement Method Of Machine Vision Based On Quaternion

The surface roughness is the microscopic geometric feature of the machined surface of mechanical parts,which is composed of the micro peaks and valleys and the distance between them.It is closely related to the mechanical properties such as fatigue resistance,corrosion resistance,wear resistance and coordination of mechanical parts.Therefore,accurate and efficient measuring the surface roughness of mechanical parts are conducive to improving the processing efficiency,and have a good promoting effect on expanding the progress of surface detection technology and related fields.Due to the advantages of machine vision,such as large area,non-contact,low cost,high detection efficiency,and automation,many scholars have applied machine vision to surface roughness measurement and achieved fruitful research results.In view of the current machine vision measuring surface roughness research present situation,the paper based on visual measuring surface roughness of quaternion method for the research of the system,through the pure quaternion modeling form of color images,the evaluation indexes based on the relative roughness of quaternion feature design research,and verify the effectiveness of the proposed method.The main research work of this paper is as follows:(1)This paper first introduces the basic knowledge of quaternion and the method of using quaternion to represent color image.According to the correlation algorithm theory of quaternion.Three algorithms,singular value decomposition of quaternion,gradient and Fourier transform,are introduced from the perspective of spatial and frequency domain algorithms,which provide theoretical support for the subsequent design and optimization of relevant roughness evaluation indexes.(2)At present,the color information characteristic index is mainly based on the subjective combination of red and green,and the coupling relationship between the space vector of the three primary colors and the roughness is still lack of mathematical profound research,so a grinding surface roughness measurement method based on the singular value entropy index of quaternion matrix is proposed.From the perspective of the mechanism of visible light imaging simulation,the data of color images taken on different levels of roughness surface are analyzed,and the quaternion singular value entropy is taken as the basis to evaluate the roughness.The experimental results show that the measurement of grinding surface roughness based on quaternion singular value entropy index is more feasible than that of color difference index,and the correlation with roughness is stronger,and the prediction result of roughness is relatively more accurate.(3)The applicability of the quaternion roughness measurement method.After theoretical analysis of the imaging of color blocks on different roughness surfaces and combined with the theory of image quality assessment algorithm,this paper proposes a method to measure the grinding surface roughness based on the quaternion gradient average index.By verifying the robustness of the evaluation index in different texture placement directions and illumination environments,it is found that the method has good applicability in different measurement environments,high measurement accuracy and certain engineering application potential.(4)At present,most of the frequency-domain indexes associated with machine vision roughness measurement are designed according to gray information,which reduces the sensitivity of evaluation indexes to roughness.In this paper,a method of surface roughness measurement based on quaternion Fourier transform is proposed.By establishing the mathematical relation model between spectrum index and roughness of color information,it is proved that the spectrum index based on quaternion is reasonable and feasible to measure the surface roughness of grinding.The experimental results show that the spectrum index based on color image has a stronger correlation with the roughness relative to the gray information index,and the prediction result is relatively accurate,which provides a broad mathematical method space for the design and optimization of the spectrum index based on color image.