Research Of Detection Algorithms For Magnetic Tile Surface Defects Images

Under the background of “Industry 4.0” and “Made In China 2025” plan, High efficiency, automation and intelligent production will be the tendency of manufacturing industry in the future. The oxysome magnetic tiles are widely used in vehicles and household appliances as an important component of electrical machines, their quality have decisive effect in the whole electric system. At present, the most magnetic tile manufacturers use manual vision inspection to detect the defects of tiles. It makes the detection cost too much and lack of stability. The innovation of digital image processing has a significant role to detective the defects of magnetic tiles.This paper adopt the new method to detect the flaws of magnetic tiles' internal and outside surface by digital image processing algorithms. The main work included in the dissertation is shown as follows:1. According to the testing index, production speed and accuracy and other technical requirements, camera, lens and other hardwares are chose. Light source and lighting system of the module of lighting, which is suitable for magnetic tile, is selected and designed through the magnetic tiles' physical characteristics. Further study on the defects' echoing characteristics and make the best lighting solution for each flaw.2. After the region of interest extraction, we proposed an efficient and ordinary noise removal algorithm with extremum & direction median filter. This algorithm uses large mask to detect noise area and small mask to restore the corrupted pixels. Simulation results show that our algorithm can remove the impulse noise efficiently with excellent detail protection and get better performance than traditional algorithms. In view of the magnetic tiles' low contrast, we propose a new gray-scale enhancement algorithm for tile images.3. Based on the characteristics after image filtering, the shortage of the traditional segmentation algorithms are analyzed. We propose two new segmentation algorithms called Defect Extraction on Magnetic Tile Surfaces Based on Lower Envelope Gray-scale Contrast and Defect Detection On Magnetic Tile Surface Based on Curve Fitting of Line Scanning. These two algorithms are complementary in segmentation and we can get the defects area accurately.4. After full analysis of the characteristics of the defect areas, we recognize the cracks from the four defects by decision tree classification firstly. Then, in consideration of the other defects' eigenvalue overlap, Fisher identification method is adopted to improve the projection mapping. Finally we obtain the four kinds of defects' category.