Ferrite Magnetic Tile Defects Detection Based on Nonsubsampled Contourlet Transform and Texture Feature Measurement

• Published in: Russian journal of nondestructive testing Vol. 56; no. 4; pp. 386 - 395
• Main Authors: Xueqin Li, Liu, Zhen, Yin, Guofu, Jiang, Honghai
• Format: Journal Article
• Language: English
• Published: Moscow Pleiades Publishing 01.04.2020; Springer Nature B.V
• Subjects: Algorithms; Characterization and Evaluation of Materials; Chemistry and Materials Science; Coefficients; Computer vision; Decomposition; Defects; Ferrites; Grinding; Image detection; Image reconstruction; Image segmentation; Materials Science; Optical Methods; Structural Materials; Texture; Tiles; image processing; magnetic tile; texture feature; defect detection; тonsubsampled Contourlet transform
• ISSN: 1061-8309; 1608-3385
• DOI: 10.1134/S1061830920040075

The ferrite magnetic tiles are widely used in industry field. At present, the defects detection in ferrite magnetic tile surfaces is done by manual work. In order to improve the defects detection efficiency and prevent missed and false detection, an automatic detection system applied to the magnetic tiles non-destructive detection was proposed based on computer vision. A suit of the automatic defects detection equipment used for magnetic tile surfaces was designed so that we can adjust the position and angle of the lightings and cameras programmatically to meet the requirement of different location for different kinds of magnetic tiles. To solve the problem of automatically detect defects from magnetic tile images which are with dark colors and low contrasts, a new hybrid algorithm which combines nonsubsampled Contourlet transform and Laws texture feature measurement was proposed to eliminate the influence of the grinding textures and extract defects. In this methodology the original image was first decomposed by nonsubsampled Contourlet transform, the characteristics of the decomposition coefficients are analyzed by Laws texture feature measurement. Then according to the texture energies of the restructured image, a segmentation threshold was determined to reset the decomposition coefficients. Finally the image was reconstructed with the reconstruction coefficients, the grinding textures were eliminated and defects were obtained by Canny operator. The experimental results show that based on the proposed method, the grinding textures can be eliminated effectively, the defects can be extracted accurately, and the accuracy rate of extraction defects can achieve 93.57%. The automatic detection system can provide an effective solution to magnetic tile defects detection industry.