RESUMO
Automated defect inspection is becoming increasingly important for advanced manufacturing. The ability to automatically inspect for critical defects early in the production cycle can reduce production costs and resources on unnecessary manufacturing steps. While there are many inspection techniques available, samples from early in a production workflow can prove challenging as they may still have systematic tooling marks, causing preferential scattering and hindering defect extraction. We propose a new imaging technique that exploits the preferential scattering from a technical surface to generate predictable fringe patterns on the sample's surface using only an array of LEDs. The patterns from this adapted fringe projection technique are imaged, and phase shifting algorithms are used to recover surface undulations on the sample. We implement this technique in the context of Hard Disk Drive platters that exhibit tooling marks from the lapping process and show that it is possible to image both highly scattering pits and scratches, as well as slow surface undulations with the same apparatus.
