Machine vision systems for measurement and inspection are metrology devices that use digital cameras, dedicated lighting and image processing algorithms to acquire dimensional and qualitative information about industrial components without any physical contact. They make it possible to inspect hundreds of parts per minute, to detect defects automatically, to take 2D and 3D dimensional measurements, to read codes and to guide robots. They are by now a key technology of in-line quality control and of Industry 4.0 automation.
Would you like help choosing the right product ?
Architecture of a vision system
A typical vision system consists of one or more cameras (CCD or CMOS, monochrome or color, with resolutions from a few MP to tens of MP), optics suited to the scene (telecentric lenses for high-precision measurement, macro lenses, standard objectives), a lighting system (diffuse, ring, dome, backlight, structured-light or polarized LED lighting) and a processing unit (embedded smart camera, industrial PC or dedicated system) running image processing software. Synchronization with the process trigger allows images to be acquired on moving parts.
Algorithms and functions
The most widely used image processing functions are edge detection with sub-pixel accuracy (for dimensional measurement), pattern matching (recognition and location of known geometries), blob analysis (analysis of connected areas for defect counting), OCR/OCV for character recognition, color matching for chromatic control, and morphological analysis. The most advanced systems integrate deep learning and convolutional neural networks for complex qualitative inspections that are hard to code with classical algorithms.
Applications
Vision systems are used for the dimensional inspection of metal, plastic and electronic components, for checking the completeness and correctness of assemblies, for verifying labels and codes, for controlling print and packaging, for inspecting caps, capsules and pharmaceutical containers, for measuring plastic molding geometries, for verifying laser welds, for inspecting SMD components on PCBs, and for guiding robots in pick-and-place operations. Smart cameras in particular offer compact standalone solutions for assembly lines.
Standards and classification
Vision systems used as measuring instruments are covered by VDI/VDE 2617 and VDI/VDE 2634 (3D optical measuring systems), ISO 10360-7 (CMMs with optical probes) and ISO 14406 (characterization of optomechanical systems). Calibration uses precision targets with markings certified to ISO 17025. For qualitative inspection applications, performance is verified with representative samples and MSA (Measurement System Analysis).
Selection criteria
Selecting a vision system requires an analysis of the field of view (FOV), of the resolution required (pixel size relative to the defect to be detected), of the process speed, of the lighting conditions, of integration with the PLC and the MES, and of scalability for further checks in the future. A feasibility study on real samples is recommended before the final sizing of the system.
The evolution of 3D vision systems using structured light and fringe projection now allows the dimensional inspection of complex free-form surfaces with accuracies comparable to those of touch-probe CMMs, but with far shorter acquisition times. This technology is spreading in the aerospace and medical sectors and in the inspection of 3D-printed components.
You must be logged in to post a comment.