Machine vision-based in-line metrology replaces the mechanical stylus of a traditional coordinate measuring machine with optical sensing. Instead of physically touching a workpiece at discrete coordinate points, a vision system captures image data from which dimensional features are extracted algorithmically.

The most fundamental enabling technique is focus-height measurement via image stacking: the imaging system scans a workpiece along the Z-axis while collecting a series of images at known Z positions. The Z-height at which a region of interest achieves peak focus — the "focus peak" — becomes the measured surface height. Mitutoyo's 2024 patent defines an image stack set spanning all Z-heights to be determined for multiple regions of interest, extracting a focus metric for each sub-image to derive a corrected Z-height measurement.

A complementary approach using strobed illumination and periodically modulated focus position (Mitutoyo, EP 2019) allows illumination strobe pulses to be phase-matched to a sinusoidally modulated objective lens position. This enables rapid Z-height stacks to be collected without mechanically stepping the stage — dramatically increasing measurement throughput compared to conventional step-and-capture methods.

For workpieces with low-contrast or highly reflective surfaces, standard edge detection fails. Mitutoyo's 2005 patent resolves this by training the system in a learning mode to generate a "line-intensified image" using a parameter-controlled technique, then applying linear transformation to the enhanced image and analysing local extrema corresponding to the target line features. This extends machine vision metrology to challenging materials typical in precision machining, semiconductor packaging, and metal stamping.

Illumination consistency across multi-station lines is addressed by Quality Vision International's 2014 patent, which discloses a pre-calibration process where illumination intensity data collected on one machine can be transferred to other machines of the same type — enabling consistent measurement accuracy across production lines without manual recalibration at each station. Learn more about patent landscape analytics for metrology technology.

Fly-mode continuous-motion image acquisition is the primary mechanism by which CMM-integrated vision systems eliminate the deceleration and acceleration cycles that dominate cycle time in traditional stop-and-measure CMM operation. Position encoder data is latched at image capture; a strobe flash freezes the part. This is the architectural foundation of Hexagon's Delta-robot CMM family (2015–2016).

Z-height focus stacking and focus-curve analysis provides the 3D surface measurement capability that allows vision systems to replace the contact probe for form and height measurements. Mitutoyo's 2024 optical deviation correction patent and their EP 2019 high-speed focus height patent represent the current state of the art for this approach. According to ISO metrology standards, non-contact methods must meet equivalent traceability requirements to contact CMMs for production acceptance.

Mechanical decoupling of imaging detectors from CMM carriers is an architectural prerequisite for achieving CMM-grade accuracy in a vibration-rich production floor environment. Hexagon's EP 2016 patent demonstrates this by mounting imaging detectors to a stationary metrology table, entirely separate from the moving carrier mechanics.

Blur-constrained program portability is a practical enabler for multi-machine production deployments. By deriving acquisition parameters from blur constraints rather than hardware specifications, the same validated inspection program executes on any machine in a fleet with consistent accuracy — critical for high-volume lines with parallel or successor inspection machines.

Machine learning inference extends in-line metrology to feature scales and measurement types physically inaccessible to contact CMMs — tens-of-nanometre semiconductor features, real-time CMP film thickness, and virtual metrology from synthetic microscopy images. For enterprise R&D teams, PatSnap life sciences intelligence and materials science analytics offer adjacent landscape coverage.

Workpiece self-identification and automatic task routing without operator intervention is necessary for realising 100% in-line inspection in high-mix, high-volume production rather than statistical sampling. Suzhou Hande's CN 2020 patent claims micrometer-level precision from this unified coordinate approach. The European Patent Office has seen accelerating filings in this sub-class over the past five years.