Research On The Surface Roughness Inspection Method Of Workpiece Based On Machine Vision

In machining,the surface roughness of a workpiece is the main indicator of surface quality and can directly affect the mechanical performance,safety and life of the workpiece,especially for workpieces with specific functions(e.g.sealing,relative movement,etc.).Currently,most of the surface roughness inspection techniques use contact gauges or optical instruments,but they cannot be applied in the actual manufacturing process because this can easily lead to damage of the workpiece surface or due to the high mechanical equipment and environmental requirements in the actual manufacturing process.By applying machine vision technology to the inspection of workpiece surface roughness,the problems of low inspection efficiency,poor accuracy and high inspection cost can be effectively addressed.In this paper,based on machine vision theory,a texture feature extraction method is used for non-contact nondestructive surface roughness inspection of end mill workpieces.The main research contents of this paper are:(1)A detailed discussion of the definition of surface roughness of the workpiece and its variability with corrugation and surface shape error;an introduction to the evaluation parameters of surface roughness according to international as well as national standards,and the selection methods of sampling and evaluation length are introduced;(2)Taking the end mill workpiece as the research object,the surface images of the workpiece are obtained and the data set is established.Based on the preprocessing of the workpiece images,features in the surface image are extracted and the model relationship between the parameters and the surface roughness is constructed;(3)To address the problems of extraction error and poor rotation invariance of traditional Gray Level Co-Occurrence Matrix(GLCM),this paper proposes an improved Gray Level CoOccurrence Matrix feature extraction algorithm and verifies its robustness than traditional GLCM through experiments;to address the problem that GLCM obtains global features of workpiece images and lacks the description of workpiece.The problem that GLCM obtains the global features of workpiece image and lacks the local description of the workpiece,this paper proposes a roughness detection method by fusing Gabor features on the workpiece surface,the dimensionality of the feature vector can be effectively reduced and the characterization ability of the features is improved;(4)Construction of Support Vector Regression(SVR)model based on vertical milling workpiece surface roughness Ra.With the eight texture features extracted based on ET-GLCM and the fused features based on Gabor as the model input and the corresponding surface roughness Ra as the output,the SVR regression model is constructed to achieve the nonlinear detection of the surface roughness of the end mill workpiece.In order to improve the accuracy and speed of surface roughness prediction,the kernel penalty factor of SVR kernel function parameters is optimized by IPSO,which avoids the problem of local minimums and the blindness of SVR model parameter selection;Through comparison with other regression models and prediction experiments on different feature combinations,it is verified that the proposed model has a superior prediction performance and the proposed texture feature extraction method is more superior.The experiments prove that it effectively solves the current problems in non-contact detection of workpiece roughness,and is also very important for the realization of other engineering applications that need to detect smoothness.