From sorting out faults to finding errors before they occur
For a long time quality inspection just meant: sorting out bad parts. When the first vision inspection systems inspected plastic packages in the 1990’s, the upmost goal was to find the defective ones. But it soon turned out that this was not the whole task: A high “false alarm rate” could easily make the systems expensive and lead to a lower productivity as they would also reject parts without recognizable errors which could still have been sold easily. Especially when decisions were made upon a thin data base the amount of falsely rejected articles could go up to several percent. The conclusion drawn for vision inspections was the same as in real life: The more raw data one has the better decisions can be made. While in the 1990’s computational power and the resolution of cameras were limited, today a well-designed vision system can easily handle the data of dozens of color cameras with megapixel resolution.