Stereo 3D Measurement on a Handheld Videoscope: From "See the Defect" to "Size the Defect"

The Gap Between Detection and Disposition

Every visual inspection ultimately answers two questions: is something there? and is it acceptable? Conventional videoscopes answer the first question well. The second — the disposition decision that determines whether a turbine blade keeps flying, a weld stays in service, or a gearbox is torn down — requires numbers. A crack-like indication means one thing at 0.2 mm deep and something entirely different at 2.0 mm. Until recently, obtaining those numbers meant removing the component to a metrology lab or mobilizing a dedicated measurement platform. Binocular stereo measurement, now available on the OmniScope C68 premium configuration, closes that gap at the point of inspection.

How Binocular Stereo Measurement Works

The principle is the same one your two eyes use. A stereo probe tip carries two matched optical channels separated by a fixed baseline. Each captures the same surface from a slightly different angle, and the processor identifies corresponding points in the two images — the disparity. Because the baseline and the optical geometry are known and factory-calibrated, disparity converts directly into distance through triangulation. The result is a dense three-dimensional point cloud of the viewed surface, generated live while the inspector watches.

Three engineering details determine whether such a system is a measurement instrument or a marketing feature:

• Baseline stability. The two optical channels must hold their relative geometry through articulation cycles, temperature swings, and the mechanical shock of field use. Any drift appears directly as measurement error.
• Calibration traceability. The conversion from disparity to distance is only as good as the reference artifacts used to establish it. Sugia’s stereo systems are factory-calibrated against traceable references across the working-distance range.
• Surface texture. Stereo matching needs visible texture to find corresponding points. Mirror-polished or featureless surfaces defeat the algorithm — an operational constraint inspectors must understand, not a defect in the instrument.

Nine Measurement Methods, One Decision Framework

The C68’s premium 3D system implements nine measurement methods: Length (point-to-point), Vertical (point-to-line offset), Depth (from a user-defined reference plane), Multisegment (cumulative path length), Area (bounded region), Missing Corner (chamfer/blend loss on edges), Sectional (profile along a section line), Circle (diameter and radius of circular features), and Gap (clearance between adjacent components, such as gearbox-to-blade spacing). Measurement resolution is 0.01 mm with error held to ≤5% of reading.

These nine map directly onto the disposition criteria inspectors actually apply. Turbine-blade tip clearance is a Gap measurement. Compressor-blade leading-edge erosion is Depth and Area. Weld-root concavity is a Sectional profile. Casting porosity is Area and Depth. The inspector’s job is selecting the method that matches the acceptance criterion — the instrument handles the geometry.

What “≤5% Error” Buys You in Practice

Consider a fitness-for-service call on a corrosion pit measured at 4.0 mm deep against a 5.0 mm reject threshold. A measurement system with ±50% error makes that reading worthless — the true value could be anywhere from 2 to 6 mm. At ≤5% error, the reading bounds the true depth to 3.8–4.2 mm: a defensible, documentable accept decision with margin. This is the difference between a videoscope that informs a decision and one that merely suggests further disassembly.

Operational Guidance for Stereo Measurement

Field experience with stereo systems converges on four practices:

1. Square up to the surface. Oblique viewing angles compress disparity and inflate error. Use the articulation to bring the optical axis as close to normal as the cavity allows.
2. Respect the working-distance window. Stereo accuracy is specified across a defined range; verify your standoff before recording, not after.
3. Light evenly. Mixed shadows create false texture. Diffuse, frontal illumination produces the most reliable point clouds.
4. Document the reference plane. For Depth measurements, record where the reference was placed. An undocumented reference makes a good number unverifiable.

The OmniScope C68 premium configuration brings this capability to a 2.85 kg handheld host with interchangeable probes from 0.95 mm to 8.0 mm — stereo measurement is supported on tubes of 4 mm diameter and above. Full specifications are on the C68 product page.