Aluminum Surface Coating Quality Of Image Inspection And Research

Texture is the basic feature of the image and an effective description of the intrinsic properties of the surface, provides important features for image understanding and classification of information. It is an important research topic in computer vision and pattern recognition. Aiming at the behind status of aluminum surface coating quality manual inspection, to carry out related research based on texture analysis.This paper analyzes the gray level co-occurrence matrix, the autocorrelation function and so on the basis of the feature extraction method, mixed the mathematical morphology and particle analysis combining Gabor filter feature extraction method. The method used parallel computing to remove unnecessary interference structure while maintaining the basic characteristics of the image and improve the texture of the aluminum surface feature extraction speed. For aluminum surface coating image classification accuracy, conducting nuclear function classification accuracy test, and then refer to the comparison test results of four kernel function classification accuracy, choosed a higher classification accuracy of radial basis function for this article. And through cross-validation experiments to determine the penalty factor and the width of the Gaussian function, to avoid the traditional experience of arbitrary choice. For texture classification aluminum surface coating image classification accuracy, traditional C- means clustering and fuzzy C- means clustering is probably between 87%-88% and fuzzy nuclear clustering classification accuracy around89.5%, and the paper proposed improvemented fuzzy kernel clustering method classification accuracy was 91.6%, improving the accuracy of classifiers.Experiments show that the proposed based texture analysis of aluminum coating surface quality inspection program is feasible, and can solve the aluminum surface coating quality inspection problems.