Abstract:
This thesis presents the image processing applications in aluminum alloy wheel inspection. The 6 defect parameters used for classification are size, gray level, hollow, grouping, alignment and contrast. The important algorithms in this thesis are the repeating image acquisition for the random noise reduction, the image compensation for the fix noise reduction, the seed growing for the segmentation and the multi thresholds for the size classification. 603 objects from 30 images are used for creating the defect parameters standard value. They are divided in 4 groups: 293 structure objects, 104 big defect objects, 119 small defect objects and 87 other objects. Another 505 objects from 25 images are used for testing the image processing efficiency. The testing objects are 239 structure objects, 102 big defect objects, 105 small defect objects and 59 other objects. From the tests on Pentium II 400 MHz, it can be concluded that the hit rate of the defect inspection is 78.15% and average processing time is 45 seconds/picture.