Image processing is often used in industrial applications. One of the most common uses for image processing is for automated visual inspection of a product before being packaged and shipped to the customer. Using these types of systems in production facilities aids the avoidance of situations where a faulty or sub-standard product is shipped to the customer.
For an example scenario, a Coca-Cola bottling plant production line was used to demonstate how image processing could be used for an automated visual inspection application. A set of images were provided, taken under near constant lighting conditions in the factory. These set of images include both good and bad bottles which the bottling company expects you to use to detect certain faults as part of the filling, capping and labelling operations before the final packaging stage.
The list of assumptions made at the start of the project:
- Constant lighting conditions
- Constant bottle positioning
- Only faults in centre bottle are required to be detected
- Faults on side bottles will be dealt with separately
The list of possible faults which can occur for the center bottle:
- Bottle Cap Missing
- Bottle Deformed
- Bottle Missing
- Bottle Overfilled
- Bottle Underfilled
- Label Missing
- Label Not Printed
- Label Not Stright
Classification results for set of images for each individual fault:
| Fault Type | # Images | # Faults Detected | Classification % |
|---|---|---|---|
| Bottle Cap Missing | 10 | 10 | 100% |
| Bottle Deformed | 10 | 9 | 90% |
| Bottle Missing | 11 | 11 | 100% |
| Bottle Overfilled | 10 | 10 | 100% |
| Bottle Underfilled | 10 | 10 | 100% |
| Label Missing | 10 | 10 | 100% |
| Label Not Printed | 10 | 10 | 100% |
| Label Not Stright | 10 | 10 | 100% |
| Multiple Faults | 10 | 9 | 90% |
Classification results for set of all 141 images:
| Fault Type | # Images | # Faults Detected | Classification % |
|---|---|---|---|
| All | 141 | 139 | 98.58% |