Structured light for 3D profiling of optical moulds and inserts
Optical moulds and precision inserts rarely reveal a defect with an obvious crack. Instead, a few micrometres of waviness, a damaged radius or a shifted parting feature can change how light travels through a lens, how a mould releases, or how an insert seats in its holder. Structured light turns that subtle geometry into a dense 3D profile, allowing a camera and projector to compare the manufactured surface with its nominal CAD model without touching the part.
For Dalian Optical Technology Co., Ltd., whose portfolio includes machine-vision monitoring and optical surface-profile testing, this approach links metrology with production quality records. For teams in Australia, it is especially useful where short-run tooling, imported equipment and demanding turnaround times leave little room for manual rework. Melbourne and Sydney support specialist moulding and medical-device workshops, while Adelaide has a strong advanced-manufacturing base. A practical system therefore needs stable calibration, clear pass/fail rules, understandable software and local support when a result is disputed.
Why structured light suits precision mould surfaces
A non-contact optical scan is valuable because mould inserts are often small, expensive and finished to tight tolerances. The process projects a sequence of coded or sinusoidal light patterns onto the surface. A camera records how each pattern is displaced, and software triangulates the visible points to create a coordinate cloud. Unlike a single photograph, the result contains height as well as lateral position, so slight curvature changes become measurable rather than merely visible.
Full-field profiling is well suited to freeform cavities, aspheric contours and mould faces carrying fine texture. It captures a broad region in one acquisition and can reveal waviness across an area that might be hidden between isolated probe points. The method also leaves the insert free of scratches from contact measurement, which matters when the inspected face will later mould a high-quality optical component.
The visual task differs from symbol recognition, yet both rely on repeatable cues. In randomised card games, players interpret ranks and suits according to consistent visual conventions; on a mould, every change in a fringe or edge is interpreted against calibrated geometry. The important advantage is not merely seeing a surface, but attaching uncertainty and scale to each point. That converts a complex, reflective object into data suitable for sorting, trend analysis and process control.
From projected pattern to height map
At the heart of the instrument is a calibrated projector-camera geometry. The projector’s pixel grid is mapped to rays in space, while the camera records those rays through the deformation of the pattern. The baseline between the two devices and the local ray angles determine depth. Calibration should therefore cover lens distortion, focus, camera pose and projector geometry, not just a ruler placed in the field. A traceable step gauge or sphere can confirm that reported coordinates remain correct after transport, vibration or a lens change.
Many systems use phase-shifting patterns. Four phase steps can produce a wrapped phase value for each pixel; a Gray-code pattern or an additional wavelength can then identify the correct order of adjacent fringe periods. Triangulation converts phase into depth, and camera calibration converts pixel position into X and Y. The finished surface may be represented as a point cloud, a regular height map or both. The representation should preserve the original measurements, because aggressive averaging can make a mould appear smoother than it is.
Surface finish can complicate every stage of this calculation. Glossy metal, black polymer and transparent coatings may return bright highlights or weak contrast, while sharp edges can mix pixels from two planes. Cross-polarisation, exposure bracketing, a longer wavelength or several viewing directions can reduce these errors. If an insert or moulded component carries an anti-reflective layer, lens coating selection also affects the optical response; coating colour, thickness and angle should be recorded with the scan so the measurement can be interpreted correctly.
Preparing moulds and configuring the scan
Preparation starts with traceability and a repeatable presentation of the part. Clean the insert with a method approved for its coating, inspect it for burrs, and secure it without loading the measurement face. Establish datums from the holder or design model rather than from whichever edge happens to face the camera. Mask the fixture, parting line and inaccessible regions so the reconstruction does not invent geometry there. Record temperature, part identifier and calibration file, especially when a scan may later support a supplier dispute.
| Method | Information captured | Main strength | Main constraint | Typical role |
|---|---|---|---|---|
| Structured light | Full-field 3D coordinates and height | Rapid coverage of curved surfaces | Sensitive to glare, edges and ambient light | Mould cavities, freeform surfaces and inserts |
| Laser triangulation | Height along a scanned line | High local depth sensitivity | Slower across large areas and sensitive to alignment | Narrow grooves, edges and datum features |
| Chromatic confocal | Fine axial height information | Detects very small surface variations | Limited field and specialised cost | Small optical surfaces and coatings |
| Contact stylus | Three-dimensional coordinates at probe points | Direct and familiar measurement | Slow, contact forces and restricted access | Datum checks and confined pockets |
| Optical interferometry | Surface height and waviness | Exceptional sensitivity on smooth surfaces | Demanding cleanliness and optical conditions | Polished reference surfaces and precision optics |
Each technique answers a different question. Structured light is efficient when an entire freeform area must be mapped. A stylus remains useful for a quick datum check or a small pocket, while interferometry provides exceptional sensitivity on smooth, reflective optical surfaces. Choice should follow the defect, material and required coverage rather than a single resolution figure, since repeatability, edge response and calibration uncertainty often matter more than the smallest nominal feature.
Once acquired, the point cloud is aligned to CAD using the same datum strategy used on the shop floor. Best-fit, datum and minimised-residual options can produce different answers, so the engineering drawing must state how the comparison is to be made. Colour maps then show positive and negative deviation, while sectional profiles expose waviness that a global tolerance might conceal. Report peak deviation, RMS error, coverage and repeatability separately. A narrow numerical tolerance with sparse coverage is not equivalent to a complete profile.
Software, tolerances and production decisions
Production software should keep the raw cloud, calibration, settings and derived report together. If the nominal CAD revision changes, the old measurement should not silently appear on a new report. Version control also allows a tool-room team to compare scans taken before and after a corrective grind. Export formats should preserve scale and units in millimetres or micrometres, matching the metric conventions used by Australian workshops, and reports should identify any interpolation, smoothing or filtering applied.
Measurement capability must be demonstrated rather than inferred from software resolution. Repeated scans of the same feature, preferably at different times and orientations, reveal repeatability; known artefacts establish bias. NATA-accredited calibration and verification services can provide an Australian traceability path for gauges used in contractual acceptance. Teams can also apply AS/NZS ISO 9001 controls to inspection equipment, maintenance and operator training, while using the relevant drawing and geometrical product specification to define datum and tolerance decisions.
Statistical process control can then connect geometry to manufacturing variables. A gradual drift in cavity curvature may precede a visible defect, while step changes can coincide with a new resin batch, insert coating or maintenance intervention. Control charts should use features tied to function, such as sagittal error or local slope near an optical zone, rather than thousands of correlated cloud points treated as independent observations. Exception rules can trigger a containment check while an engineer investigates the root cause.
Controlling variation in Australian production environments
Australian production environments add practical constraints that should shape the installation. Melbourne’s variable humidity can shift polymer dimensions and make contamination visible on polished surfaces, Sydney’s coastal salt and airborne dust can degrade optical paths, and Adelaide’s summer heat can warm a camera or distort a fixture. Air-conditioning the measurement cell, controlling stray light and allowing the equipment to reach thermal stability are routine requirements, not optional refinements. A fixed scan recipe for every site ignores those local realities.
Workplace duties also apply to the complete inspection cell. Under the model Work Health and Safety Act 2011 and state-specific electrical and laser-safety rules, employers must manage electrical testing, interlocks, emergency stops and training, plus laser controls where a laser projector is used. Risk assessment should consider projector light, moving stages and ergonomic work, even when the device itself is enclosed. Software that stores customer drawings or production records should be covered by appropriate access controls and, where applicable, obligations under the Privacy Act 1988.
Human factors matter as much as ambient conditions. Australian tool rooms often rely on small, multi-skilled teams, so the operator needs a short, repeatable workflow and results that can be explained without specialist software knowledge. Heat stress is a familiar concern during summer loading, and workers should follow workplace heat-management procedures. Anyone prescribed propranolol medication should seek advice from their clinician and follow the label, because medication can affect an individual’s response to heat or exertion; operational controls must never depend on workers self-selecting out of hot tasks.
Turning profile data into process improvements
The useful endpoint is not a coloured cloud on a screen but a closed loop between measurement and production. A repeatable deviation map can reveal whether a mould is drifting, whether polishing is removing too much material, or whether an insert is mounted at the wrong angle. Combining geometry with cycle time, cavity pressure, temperature and inspection results helps separate a true tooling problem from contamination or fixture error. The same dataset can support incoming supplier checks, first-article approval and scheduled condition monitoring.
Optical inspection has already demonstrated value in related production settings. An eyewear inspection case study illustrates how structured measurement can reduce manual sorting and make process feedback faster, although an optical mould programme needs additional attention to datum definition and coating response. The lesson is to define the decision each measurement must trigger: rework, quarantine, engineering review or routine release. A report with no agreed response adds data but little control.
Dalian Optical Technology Co., Ltd.’s broader portfolio in machine-vision monitoring, lens inspection and related software can support this workflow, while integrators should pair it with local calibration, installation and application engineering. For Australian buyers, the strongest implementation is a validated cell rather than a standalone camera: calibrated geometry, controlled conditions, documented recipes and clear escalation paths. When those elements are aligned, 3D profiling becomes an objective production signal that protects optical performance and reduces avoidable rework.