Research On Segmentation And Inspection Of Magnetic Tile Defect Image

With the advancement of science and technology,the permanent magnet motor is gradually integrated into human life.As its essential part,the quality of the magnet tile needs to be strictly controlled.Therefore,a series of methods are proposed to realize the segmentation and detection of the surface defect of the magnet tile.This thesis is based on the machine vision theory.Based on the existing detection and recognition technology,a technical algorithm suitable for the segmentation and detection of magnetic tile surface defects is selected,and the corresponding algorithm is improved to realize the segmentation and detection of magnetic tile surface defects.The specific work is as follows:(1)Aiming at the problem that the collected magnetic tile image needs to reduce noise,this research will perform quantization processing,Gaussian filtering,gamma correction and other operations on the original image to enhance the quality and contrast of the original image and provide a good foundation for subsequent image segmentation;(2)According to the Otsu method,the segmentation results of the magnetic tile image are analyzed,the concept of fuzzy clustering is introduced,the intuition factor is introduced into the fuzzy clustering algorithm,and the influence parameters of the non-membership matrix are added to the original clustering objective function.Describes the mutual constraint relationship between the center pixel and adjacent pixels,thereby improving the clustering performance of the image target area,overcoming the noise and shadow interference in the magnetic tile image,using the CH clustering index to determine the number of clusters,and finally achieving image segmentation.The results show that the improved fuzzy clustering effectively separates the target area and the background,and has superiority in the time dimension;(3)Extracting the morphological and texture features of defects,extracting 14 features in total,using principal component analysis technology to process multiple features,selecting the top 6 features as model input parameters;constructing a support vector machine multi-classification model,using grids The search selects the optimal penalty parameter and gamma function to realize the defect detection of the magnetic tile image.Compared with the KNN model,the SVM model finally wins with an overall recognition rate of 84.4793%.Experiments show that the support vector machine has good performance in terms of accuracy and robustness,and can effectively detect cracks,pores,and collapse defects on the surface of the magnetic tile.