Defects detection on bottles is one of “PAWLIN Technologies” projects which is under preliminary research for one european organization.
There was developed an experimental sample for detecting such defects as spots, cracks, etc.
While creating training sample the bottle is photographed opposite light sources. Training mode is intended to create a signature of a reference object (without defects) based on a set of object images.
If any defect detected, the test image signature is extracted. Then using the MSER algorithm of cluster matching, the test signature is being compared with the images signature of a reference object. During this process the algorithm calculates the estimate deficiency of an object and decides whether the object is defected or not based on the threshold decision rule.
“PAWLIN Technologies” is looking for cooperation with companies which a interested in integrating automatic production control solution on their conveyor production systems.