VDI/VDE/VDMA 2632 Blatt 3:2017-10

Quantitative capability considerations are already standard for measuring (not classificatory) systems. Measurement uncertainty is normally used here as a parameter. There are established procedures as to how this can be calculated and how suitability for a given task can be deduced from this. For classificatory machine vision systems (MV systems) whose results are attributive variables, so far no corresponding established parameters exist. This standard introduces characteristics describing the classificatory performance of a machine vision system. Due to the variety of forms taken by classificatory MV systems it is scarcely possible to provide a general scheme for assessing the performance of any particular systems. The standard suggests procedures for assessing classificatory performance during the acceptance process for MV systems. Typical examples of testing techniques for industrial imaging systems are given that may serve as guidance for similar cases. It is thus addressed to both users and suppliers of MV systems.