Advanced Inspection and Measurement (AIM) has developed a number of defect inspection capabilities for translucent materials. Data is acquired utilizing line or area array cameras capturing images under front or backlit conditions. Cylindrical part circumference surfaces are typically imaged by rotational scanning utilizing line scan cameras, flat surfaces are typically imaged by area array cameras. Backlighting provides the ability to detect defects within the material as well as on the surface, while front lighting allows for surface defect detection only. Part handling / packaging can be provided to any level to suit the customers needs. Data storage and archiving can be provided to customer specifications.
A typical acquired raw image contains a great deal of contrast variation due to part shape, density and position. Applying a direct threshold to this image to define the defect areas would not work to isolate all of the unique defect areas.
The raw image can be normalized to minimize the variation of contrast due to part shape, density and position. Applying a threshold to the normalized image yields defect clusters, which need to be further, evaluated to determine if they are cause for rejecting the part.
The identified defect clusters are evaluated and assigned a value for a variety of attributes to aid in classification of the defect. A typical attribute list is provided below:
| Area |
Length |
| Density |
Aspect Ratio |
| Concavity |
Perimeter |
| Symmetry |
Diffusion |
The defects can now be classified according to the highest score produced by the weighed attribute sums for each defect classification. The classification yielding the highest score will become the classification for the particular defect. Typical defect classifications are listed below:
- Flash
- Scratches/Chips
- Cracks
- Void/Bubble
- Spot
The classified defects can now be compared to individual reject size limits to determine an overall accept / reject condition for the part.
Copyright ©2000-2006 Advanced Inspection and Measurement, Inc. All rights reserved
