Robustness in machine vision systems
Nowadays, everyone is aware of the importance of machine vision systems for production quality assurance. For this reason, the use of vision systems is widespread and continues to expand.
A very important point is validation.
Once the system is installed and configured, a complete validation has to be carried out in order to put the system into production with full reliability.
Validation often complicates the start-up, as a significant sample of images is needed with all the situations that can be found in production, variability of lighting in the work area, variability in the position/orientation/finishing/colours... of the part to be inspected.
Hence the importance of automating this whole process and thus being able to carry out the validation as efficiently as possible and also to be able to schedule periodic automatic validations to verify that the system is not degraded and that the new small variations, which usually occur, are absorbed without any problems.
Apart from purely vision issues, it is no less important to check that there are no problems with cameras, lights and the different applications involved in the process. For example, if a camera remains in an unknown state and always sends the same image to the computer, this would represent a serious fault due to lack of control.