Machine Vision for Polished Surface Roughness
Machine vision for measuring surface roughness of polished optical parts combines calibrated cameras, controlled illumination and analytical software to reveal finish defects that may be invisible to the eye. Microscopic scratches, coating orange peel, tool marks and residual grinding texture can scatter light, reduce transmission and create artefacts in high-performance imaging. The objective is not simply to photograph a shiny component, but to turn its surface response into a repeatable quality signal linked to engineering limits.
For Australian optical, medical-device and advanced-manufacturing operations, this approach offers rapid full-field screening alongside laboratory profilometry. It is especially useful when polished lenses, prisms, windows and reflective components pass through production in batches. The strongest implementation combines an optical sensor, stable fixturing, suitable lighting, reference samples and a controlled software workflow, while acknowledging that a camera image alone cannot provide traceable roughness values.
What machine vision can measure
Surface roughness and surface profile are related but not interchangeable. Roughness usually describes fine-scale texture, while a surface profile can also include longer-wavelength waviness and the component’s overall form. Common parameters include arithmetic average roughness, Ra; root-mean-square roughness, Rq; and maximum height, Rz. Parameters such as these are defined in standards including ISO 4287 and ISO 25178, although those standards define how results are reported rather than prescribing a particular camera-based measurement method.
Automated imaging can detect roughness through changes in reflected or scattered light. Dark-field illumination suppresses the main specular reflection and makes small surface irregularities appear as bright features. Bright-field imaging records reflectance differences caused by scratches, coating defects or polishing marks. More sophisticated phase-sensitive systems can interpret fringe patterns and reconstruct height variation. To produce a calibrated Ra or Rq result from camera data, the relationship between image features and a physical reference measurement must be established for the relevant material, coating, geometry and production process.
The approach suits components where appearance, stray light and coating integrity affect performance. Typical examples include lens and eyewear elements, moulded optical surfaces, precision windows and reflective solar-thermal components. The product range associated with Dalian Optical Technology Co., Ltd. covers optical sensors, machine-vision monitoring systems, lens and eyewear inspection equipment, optical surface-profile testing and related software, providing several possible layers within the same quality-control workflow.
Comparing optical measurement approaches
No single optical method handles every finish requirement. The appropriate choice depends on whether the operation needs defect detection, comparative ranking, process monitoring or quantified roughness values.
| Method | Signal captured | Best suited to | Main limitation |
|---|---|---|---|
| Dark-field scatter imaging | Spatial distribution of light scattered by surface texture | Fast screening, defect detection and calibrated roughness ranking | Strongly affected by lighting, material and part geometry |
| Bright-field reflectance imaging | Changes in reflected intensity and appearance | Coating uniformity, scratches, tool marks and visible finish defects | A polished surface can hide texture, and intensity is not a direct height measurement |
| Confocal or focus-variation profilometry | Three-dimensional height data | Quantitative Ra, Rq, Rz and detailed topography | Greater cost, slower acquisition and possible limits on steep or reflective surfaces |
| Interferometric measurement | Fringe phase converted into surface height | High-sensitivity roughness and waviness analysis | Vulnerable to vibration, reflections and surface-coating conditions |
Scattering and reflectance imaging are valuable when production speed and full-field coverage matter. They can flag parts according to validated limits and provide useful process trends without assigning a precise physical value to every pixel. This makes them particularly effective for in-line inspection of polished optical components.
Quantitative profilometry or interferometry is preferable where a drawing specifies Ra, Rq or another dimensional surface-texture parameter. A practical system often uses both: the camera rejects obvious defects and monitors process drift, while a smaller group of parts is checked against a reference instrument. A consistent correlation study can then define when camera measurements may be used for routine release decisions.
Designing a stable inspection cell
Camera selection begins with the smallest defect that must be detected and the working distance available. Monochrome cameras often provide greater sensitivity and spatial consistency than colour sensors when the inspection depends on subtle intensity variation. Telecentric optics can reduce magnification changes near the edge of the field, while optimised lenses and filters help preserve contrast on highly reflective parts. Resolution should be assessed at the actual part surface rather than from sensor-pixel specifications alone.
Lighting and mechanics determine whether the captured signal is repeatable. A fixed-angle dark-field illuminator, diffuse dome or polarised bright-field source may suit different components, but the configuration must match the finish being inspected. Consistent part orientation, non-contact fixturing and repeatable focus are equally important. Small variations in illumination angle or working distance can alter the apparent severity of polishing marks, so automatic exposure, camera settings and calibration checks should be controlled through the inspection software.
Software must translate pixels into manufacturing information. It should distinguish isolated scratches from broad coating texture, remove known edges and fixturing features, and compare results with an approved acceptance model. Statistical process-control functions can chart Ra proxy values, defect counts and regional trends by machine, batch, shift or polishing process. Before choosing the sensor and analysis settings, teams should review surface profile testing to clarify the intended measurement principle and the difference between visual screening and dimensional surface analysis.
Deployment considerations for Australian manufacturers
Australian buyers should match the inspection method to the local industry using the component. Sydney and Melbourne have established photonics, medical-device and precision-engineering sectors, while Brisbane supports aerospace and advanced manufacturing. Adelaide’s defence and space activities create requirements for reliable optical-component inspection, and Perth’s energy and mining sectors use reflective equipment and instrumentation that can benefit from automated finish monitoring. These applications differ in tolerance, throughput and environmental exposure, so an Australian installation should be demonstrated on representative parts.
Importing equipment from Dalian, Liaoning also requires normal Australian customs and procurement planning. The importer should obtain the correct tariff classification for the complete machine, confirm any applicable China–Australia Free Trade Agreement treatment and check GST, freight, insurance and spare-parts costs with a customs broker. Zero duty should not be assumed until the goods’ origin and classification have been accepted. Local electrical safety and electromagnetic-compatibility requirements may also require an RCM pathway, supplier declarations and supporting technical documents before the equipment is sold or installed.
Metrology traceability should be established independently of the camera supplier. Reference samples should be calibrated by a laboratory whose accreditation scope covers the relevant surface parameter, and the method should be checked against ISO 4287 or ISO 25178 reporting requirements where applicable. NATA accreditation is a useful reference when selecting an Australian calibration provider, but the exact scope must be confirmed. International buyers should also review company registration, current technical documents, installation requirements and test data rather than relying on web copy alone.
Practical recommendations for production control
Begin with a structured trial that links image results to known-good and known-defective parts. Measure repeatability across operators, machines, shifts and several production lots before defining automatic acceptance limits. A measurement-systems analysis can reveal whether the inspection distinguishes acceptable finish from actual defects rather than merely detecting large visible changes.
- Define the required output as defect screening, comparative roughness, Ra/Rq reporting or another measurable characteristic.
- Standardise lighting, part orientation, focus, exposure and fixturing before fine-tuning the defect algorithm.
- Correlate camera results with traceable profilometer or interferometer measurements on representative components.
- Use acceptance limits derived from customer drawings and process capability rather than universal roughness thresholds.
- Plan calibration access, spare-parts stock, software version control and local technical support before deployment.
Once validated, the inspection cell can support first-article checks, in-process monitoring and final inspection without slowing every part for contact measurement. The software can retain images and numerical features for traceability, while control charts reveal gradual changes caused by polishing-wheel wear, slurry condition, coating thickness or fixture contamination. Periodic reviews should confirm that software updates and changes in materials or production methods do not invalidate the original correlation.
For buyers evaluating equipment from Dalian Optical Technology Co., Ltd., requesting sample analyses, technical documentation and a documented trial through contact the supplier provides a sound basis for local acceptance. A successful Australian deployment links fast optical inspection with verified reference measurements, producing dependable surface-finish control while retaining clear traceability for each component.