AI/Self-Learning: NxtGen™ Intelligent Classification Engine (NICE)

Defect classification can be a time-consuming and frustrating endeavor. Conventional rule-based classification becomes untenable as defect classes proliferate; logic gaps and overlaps result.

Dark Field’s new NICE Artificial Intelligence (AI) module utilizes multiple classification algorithms including Rules-Based, Nearest Neighbor, Bayesian and Neural Network. This package of technologies:

• Provides selection of the best algorithm for the classification task.
• Utilizes an extended set of geometric, intensity, and topology features.
• Leverages human intuition in a easily trained classifier.
• Delivers robust and reliable classification in real-time.

Streak and scratch detection: Special signal processing algorithms extract weak, persistent signals from a background of noise.

Repeat defect detection: NxtGen™ processes defects in the same cross-direction position and alerts the operators, in real-time, to the presence of repeat defects, like scratches, indentations, coating skips, and more.  Digital outputs alert the operator when any repeat defect is present.

Defect density and cluster detection: Often, manufacturers will accept small defects. However, if the number of small defects in a given area exceeds a certain value, these must be flagged and rejected. NxtGen™ utilizes a “sliding window” algorithm to locate areas of excessive defects and alerts the operator with mapping, alarms and images.