Quality inspection of glass defects in a complex environment

Quality inspection of glass defects in a complex environment

Challenge


Quality problems with glass vial and syringe results in yearly losses of ~8 B$ for the pharmaceutical industry. Current automated inspection techniques only allow the inspection of filled containers, resulting in waste. Moreover, the diversity and complexity of the defaults is too high to be managed by standard algorithms.

Solution


Using a vision system and deep neural networks, our inspection system is able to detect particles larger than 25µm in un-filled containers with varying shapes and backgrounds.

Results


  • >95% contaminant detection accuracy
  • 100% inspection instead of random sampled inspection
  • Sizeable savings by reduction of waste and product recalls