Cameras for fault detection are typically referred to as Vision systems and they can get quite expensive depending on what you need. One of our vendors recently spent $500K on a system for checking of artwork on their food cans. The system scans the entire 360° of the can, using 8 cameras, looking for print/artwork defects and rejects any defects.
For just checking labels it shouldn't be too cost prohibitive. Do your labels have bar-codes? If they have a bar code then the vision system can be set just for checking that. It will still require an operator to configure/program the Vision system on the production run so it knows what bar code it's looking for. The same if you're just checking the artwork. An operator will still need to set up the vision system before the run so the system knows what artwork it's looking for.
We have in place a three person sign off on checking of product inkjet date coding and we still have failures. People look but they don't SEE.
Computerised Vision systems remove the human element by about 99.9999% of the times but they are still reliant on human input for set-up, verification and validation. But even the best vision systems are not 100% fool proof. The vendor with the can vision system - it was rejecting 100's of cans. An operator looked at the cans being rejected and couldn't find any fault with the artwork so they overrode the rejection. When we got the cans they looked OK to us as well and we didn't see the defect either until we started the filling process. The issue was that the artwork was perfect, just upside down. The vision system knew this but the operator couldn't see the defect because the artwork was actually OK, they were not looking at how it was placed on the can. Operators get so used to seeing their product they didn't see the problem.
If you haven't already, do a web search for vision systems and you may get some idea of what they can/cannot do.
