Demonstrating how image processing can be used for an automated visual inspection application using a set of images from a coca-cola production line as an example scenario.
The program makes the following assumptions:
- Near-constant factory lighting conditions
- Consistent bottle positioning
- Only faults in the centre bottle are required to be detected
- A missing centre bottle is not considered a fault, it is simply ignored
- The folder to be processed is a direct child of '.\Pictures' (Example folders have been included for testing purposes)
The list of faults that are to be detected are:
- Bottle Cap Missing
- Bottle Deformed
- Bottle Overfilled
- Bottle Underfilled
- Label Missing
- Label Not Printed
- Label Not Straight
Note: A bottle is not considered overfilled if the bottle is deformed as the overfilling is most likely due to the deformity. It will still be flagged as faulty regardless.
Results are based on the included 'All' folder consisting of 141 images, including bottles with single faults, multiple faults, and no faults.
| Folder | Number of Images | Correctly Identified | Accuracy |
|---|---|---|---|
| All | 141 | 136 | 96.45% |