Surfaces with more or less chaotic patterns arise in many manufacturing industry fields whether it represents a shortbread, a scrubbed metal surface or the porous surface of a sensor.
Such surfaces always are new challenges for an optical inspection system because the target pattern often hardly differs from a real surface defect and most of the time it is barely mathematically describable.
In the course of a doctoral thesis the dwh GmbH has developed an algorithm which automatically detects errors in chaotic patterns without requiring lots of training data. Also the adaptation to new patterns in production is possible without a learning phase.
First tests based on sensor surfaces showed promising results. Furthermore, the processing speed is suitable for real time processing.