Optical, Coordinate and Non-Contact Metrology

Introduction

Modern manufacturing increasingly requires measurement of complex, miniature, free-form and delicate components. Conventional contact gauges remain valuable, but they may be too slow, physically intrusive or geometrically limited for such work. Optical instruments, coordinate measuring machines (CMMs), machine-vision systems, laser scanners and structured-light scanners extend measurement from isolated dimensions to complete geometric descriptions.

These technologies do not eliminate metrology fundamentals. A measurement is still a comparison between a measurand and a reference, and its result is meaningful only when the instrument, datum system, environment, probing strategy, software and verification procedure are controlled. Non-contact does not mean error-free: optical reflections, shadows, calibration errors, sampling density and surface texture can strongly influence results.

This chapter develops the principles and practical use of optical and coordinate-based systems, with emphasis on alignment, probing, point-cloud analysis, verification and quality decisions.

Learning Objectives

By the end of this chapter, the learner should be able to:

1. Explain the operating principles and applications of toolmakers’ microscopes and profile projectors.
2. Describe CMM architecture, coordinate systems, probing strategies and geometric fitting.
3. Configure and interpret vision-system measurements.
4. Compare laser scanning and structured-light methods.
5. Establish datum alignment and distinguish alignment from best-fit comparison.
6. Analyse point clouds and recognise major sources of uncertainty.
7. Plan verification, calibration and traceability activities for advanced measurement systems.
8. Select a suitable method for a particular manufacturing inspection problem.

1. Optical Metrology

1.1 Toolmakers’ microscope

A toolmakers’ microscope is an optical measuring instrument in which a magnified image of a workpiece is viewed through an eyepiece, camera or digital display. The workpiece is normally mounted on an X–Y stage, while angular rotation, vertical focus and illumination are adjustable. Crosshairs, reticles or edge-detection software provide a reference against which features are located.

Typical applications include measuring small holes, slots, thread forms, cutting-tool geometry, radii, angles and distances between features. The instrument is especially useful where contact probing could deform a thin part or damage a sharp edge.

The measured distance between two points in a calibrated image is commonly determined from the stage displacement or image scale. If a feature measures n pixels and the calibrated scale is s millimetres per pixel, the image-derived length is:

Here L is the measured length in millimetres, n is the pixel separation and s is the calibrated length per pixel. The relationship assumes constant magnification and a properly calibrated, distortion-corrected image. It is not valid near an uncorrected optical field edge where scale distortion varies with position.

Important error sources include parallax, focus error, imperfect edge definition, backlash, stage squareness, lens distortion, thermal expansion and operator selection of the apparent edge. Dark-field illumination often improves the visibility of profiles, whereas bright-field illumination may be preferable for surface or internal detail.

1.2 Profile projector

A profile projector, or optical comparator, projects a magnified silhouette of a component onto a screen. The operator compares the profile with a drawing overlay, a template or a computer-generated nominal profile. Linear dimensions and angles can be measured using a movable crosshair or screen coordinate system.

The magnification is:

where M is magnification, I is image size and O is object size, with I and O expressed in the same units. For example, a 20 mm object producing a 100 mm image has M = 5. The magnification must be verified because nominal lens magnification can differ from actual scale owing to optical and mechanical errors.

Profile projectors are effective for two-dimensional features, but they do not directly establish three-dimensional location or form. A tilted part, incorrect focus plane or unsuitable illumination may cause a profile to be interpreted incorrectly. The workpiece must be securely and correctly oriented, and the screen or digital image must be calibrated with a traceable reference.

1.3 Vision systems

A machine-vision system normally consists of illumination, lens, camera, image-processing software, a fixture and sometimes a motion stage. The measurement chain is:

Figure: Functional chain of a machine-vision measurement system.

Common image operations include thresholding, filtering, blob analysis, edge detection, circle fitting and pattern matching. A threshold converts grey-level information into object and background regions. Thresholds should be validated against expected changes in lighting, material and surface finish rather than selected only for one sample.

2. Coordinate Measuring Machines

2.1 Architecture and operating principle

A CMM determines the coordinates of a stylus or optical sensor relative to a reference frame. Bridge, cantilever, gantry and horizontal-arm configurations are selected according to workpiece size, accessibility, rigidity and required uncertainty. The machine includes guideways, scales or encoders, a probing system, controller and measurement software.

A contact probe records a point when the stylus touches the workpiece. A scanning probe records many points while moving across a surface. Non-contact probes may use cameras, lasers or other optical sensors. The measured point is not always the point physically touched: stylus radius, probe qualification and approach direction must be compensated.

A measured point in three-dimensional space is represented as:

where x, y and z are coordinates in the active Cartesian frame. A rigid transformation between two frames can be written:

where R is a 3 × 3 rotation matrix satisfying RᵀR = I, t is a three-dimensional translation vector, and p_A and p_B are the same physical point expressed in frames A and B. This model assumes a rigid workpiece and negligible scale change; it is fundamental to datum alignment and registration.

2.2 Probing strategies

A probing plan should be based on the geometry, datum scheme, tolerance requirement and uncertainty budget. Touch-trigger probing is efficient for discrete points. Scanning is preferable for form, contour and profile evaluation, but it produces more data and requires suitable filtering and path control.

For a plane, three non-collinear points are the minimum mathematical requirement, but additional well-distributed points are normally necessary. For a circle in a plane, at least three non-collinear points are required for an ideal fit. For a cylinder, points should be distributed around the circumference and along its length; points concentrated in one small region can produce an unstable estimate of axis and diameter.

Probe qualification establishes the effective stylus-tip radius and probe geometry using a calibrated sphere or other reference. Qualification should be repeated when the stylus, holder, probe orientation or operating conditions change. The probe approach should avoid collisions and should be approximately normal to the local surface where practical.

2.3 Geometric fitting

Measured points rarely lie exactly on an ideal feature because of manufacturing variation, sampling error and measurement noise. Software therefore fits an ideal line, plane, circle, cylinder or other geometry. A least-squares plane can be expressed as:

with the constraint a² + b² + c² = 1. The signed perpendicular distance of point (xᵢ, yᵢ, zᵢ) from the plane is:

The least-squares fit minimises ∑rᵢ². This approach is appropriate when random errors dominate and all points should contribute according to the selected fitting rule. It is not automatically appropriate for every tolerance. For example, minimum-zone evaluation seeks the smallest separation between two parallel boundaries containing the data, while least squares seeks the best average fit. The fitting method must match the drawing standard and inspection requirement.

3. Coordinate Systems and Alignment

3.1 Machine, part and feature coordinates

A CMM has a machine coordinate system, but inspection results are usually reported in a part coordinate system established from functional datums. Alignment converts raw machine coordinates into coordinates meaningful for design and assembly.

A common datum alignment follows the 3–2–1 principle:

• Three points on the primary datum establish a plane and constrain translation normal to it and two rotations.
• Two points or a line on the secondary datum constrain one translation and one rotation.
• One point on the tertiary datum constrains the remaining translation.

The exact implementation depends on datum feature geometry and the applicable geometric tolerancing standard. A datum plane should not be created from poorly distributed or contaminated points, and a datum axis should be sampled over sufficient length to represent the functional feature.

3.2 Best-fit alignment

Best-fit alignment minimises discrepancies between measured and nominal data, often using an objective such as:

where pᵢ is a measured point, qᵢ is the corresponding nominal point, wᵢ is a weighting factor and n is the number of point pairs. The norm represents Euclidean distance. This equation assumes known correspondence and a rigid transformation; different weighting, outlier rejection and distance definitions produce different results.

Best fit is useful for analysing overall shape or matching free-form surfaces, but it can conceal functional datum errors. A part may appear excellent under a free best fit while a mounting hole is incorrectly located relative to the actual datum scheme. Therefore, datum alignment is normally used for acceptance, while best fit is used as a supplementary diagnostic unless the specification explicitly requires it.

4. Non-Contact Measurement Methods

4.1 Laser scanning

A laser scanner projects a laser line or point onto a surface and observes the reflected light from a known angle. Triangulation geometry converts image position into three-dimensional coordinates. In a simplified arrangement, if baseline B is the distance between projector and camera, focal length is f, and image disparity is d, depth z is proportional to:

Here f and d must use compatible length units, while B and z have the same length unit. This simplified equation assumes a calibrated pinhole geometry and does not represent the full camera model, lens distortion or multi-view processing used in commercial systems.

Laser scanning rapidly collects dense data and is useful for free-form surfaces, large parts, reverse engineering and comparison with CAD. Limitations include sensitivity to surface reflectivity, translucency, colour, ambient light, line-of-sight obstruction and scanner stand-off. Glossy metal may require approved removable coating, but coating thickness can itself affect results and must be controlled.

4.2 Structured light

Structured-light systems project coded patterns, commonly stripes or grids, onto a surface. One or more cameras observe pattern deformation, and triangulation reconstructs surface coordinates. Phase-shifting systems may use several patterns to improve spatial resolution, whereas coded patterns may identify correspondence over larger areas.

Structured light can acquire a complete field of view quickly, making it effective for moulds, castings, ergonomic shapes and inspection of large surfaces. It is less effective where the surface is hidden, highly reflective or outside the calibrated working volume. Multiple views must be registered, often using targets, common geometry or a controlled turntable.

4.3 Optical and non-contact comparison

5. Point-Cloud Processing

A point cloud is a set of measured coordinates, often accompanied by intensity, colour or quality information. Processing should preserve the distinction between measured data and software-generated geometry.

A defensible workflow is:

1. Define the measurand, tolerance and required reporting format.
2. Calibrate the sensor and verify the working volume.
3. Acquire overlapping data with adequate coverage and resolution.
4. Remove obvious non-part data, documenting the rule used.
5. Register separate views using targets, known features or constrained algorithms.
6. Establish the required datum or reference alignment.
7. Fit geometry using the specified mathematical rule.
8. Compare the result with nominal CAD or tolerance zones.
9. Inspect residuals, outliers, coverage and uncertainty.
10. Archive raw data, processed data, settings and the final report.

Registration may use corresponding targets, feature-based matching or iterative closest point (ICP) algorithms. ICP repeatedly associates nearby points, estimates a rigid transformation and updates the alignment. It can converge to an incorrect local solution if the initial alignment is poor or the part is geometrically repetitive. It should therefore be constrained and independently checked.

A colour map is useful for visual diagnosis, but colour alone is not a measurement result. The report should state the alignment, deviation sign convention, scale limits, filtering, mesh generation method and tolerance criteria. Excessive smoothing can hide real defects; insufficient filtering can exaggerate noise.

6. Verification, Calibration and Uncertainty

Calibration establishes the relationship between instrument indication and a reference standard. Verification determines whether the system continues to meet specified performance requirements under intended conditions. They are related but not identical: calibration may produce correction information, while verification tests conformance.

A measurement result should be considered with its uncertainty. If independent standard uncertainties are u₁, u₂, …, uₖ, a simple combined standard uncertainty is:

This root-sum-square form assumes the components are independent or uncorrelated. If correlation exists, covariance terms are required. An expanded uncertainty is commonly written:

where k is a coverage factor selected to provide the intended level of coverage under the uncertainty model. The result should be reported as measured value ± U, with the coverage basis stated.

Relevant uncertainty contributors include machine scale error, thermal expansion, probing repeatability, probe qualification, optical calibration, surface reconstruction, alignment, datum form, operator influence and environmental vibration. For a length affected by thermal expansion, the approximate change is:

where α is the coefficient of linear thermal expansion in K⁻¹, L is original length and ∆T is temperature change in kelvin or degrees Celsius difference. The approximation assumes uniform temperature and small dimensional change.

A CMM laboratory should control temperature, humidity, vibration, cleanliness and air flow. The workpiece should stabilise thermally before measurement. Verification artefacts may include calibrated spheres, step gauges, ring gauges, ball bars or calibrated reference parts, selected according to the machine and claimed performance.

7. Worked Example: Thermal Effect and Reporting

A steel component has a nominal length of 200 mm. Its coefficient of linear expansion is 11.5 × 10⁻⁶ K⁻¹, and it is measured 4 K above the reference temperature. The estimated expansion is:

Thus the component is approximately 9.2 µm longer than at the reference temperature, assuming it has reached uniform temperature. If the instrument reports 200.018 mm at the elevated temperature, a simple temperature correction to the reference condition gives approximately:

200.018 mm − 0.0092 mm = 200.0088 mm.

This correction is meaningful only if the material, temperature and reference condition are known and the instrument itself has been appropriately compensated. It should not be applied blindly to a complex part with temperature gradients.

Suppose independent standard uncertainties are 3 µm for machine and probe effects, 4 µm for temperature correction, and 2 µm for repeatability. Then:

Using k = 2, the expanded uncertainty is approximately U = 10.8 µm. A suitable statement is therefore 200.009 mm ± 0.011 mm, k = 2, subject to the stated assumptions and uncertainty model.

8. Applied Case: Inspection of a Precision Pump Housing

Background

The following is an illustrative industrial case. A manufacturer produces an aluminium pump housing containing a machined bore, mounting holes and a cast free-form sealing surface. Rejection rates increased after a tooling change, but conventional gauges could not identify whether the problem was bore position, sealing-surface distortion or casting variation.

Problem

The bore diameter and hole locations had functional datum tolerances. The sealing surface required comparison with CAD, but its dark, semi-reflective finish produced inconsistent scanner returns.

Analysis

A contact CMM was selected for the bore, datum faces and mounting holes because these features controlled assembly. A structured-light scanner was selected for the sealing surface because dense coverage was needed. The inspection plan separated acceptance alignment from diagnostic alignment.

The CMM probe was qualified with a reference sphere. The primary datum face was sampled at distributed locations, the bore established the secondary axis, and a mounting feature established the tertiary direction. The scanner was calibrated over its working volume, and the surface was cleaned. A thin, approved matte coating was applied only after a trial confirmed that its thickness was negligible relative to the tolerance or could be compensated.

Multiple scanner views were registered using reference targets. The cloud was filtered conservatively, aligned to the functional datum system, and compared with CAD. Residual plots were examined rather than relying only on a colour map.

Outcome

The CMM showed that bore diameter was acceptable but the bore axis was displaced relative to the mounting pattern. The scan showed a local sealing-surface distortion caused by clamping during a machining operation. A repeat scan after revised fixturing confirmed that the distortion decreased. The manufacturer revised the fixture, retained the CMM datum routine and used scanning as a process-diagnostic tool.

Learning points

• The fastest sensor is not necessarily the correct acceptance instrument.
• Functional datums must govern pass/fail decisions.
• Dense point clouds reveal patterns but require controlled registration and filtering.
• Independent methods can separate feature-location error from free-form surface error.
• Measurement planning should address the manufacturing process, not merely produce a colour image.

9. Limitations, Governance and Contemporary Practice

Optical systems can encode bias through lighting, threshold selection, lens distortion and automated edge rules. AI-assisted segmentation and anomaly detection can improve throughput, but a model trained on limited surfaces may fail on new finishes, contamination or unusual defects. Any AI-supported measurement decision should retain raw images or clouds, software version, model version, confidence information and human review rules.

Cybersecurity is relevant when CMMs and scanners connect to production networks. Controlled access, software validation, backups and change records protect measurement programs and inspection history. Personally identifiable information is rarely central to dimensional metrology, but factory images may still reveal proprietary products or processes and should be governed accordingly.

Standards-based practice should use the applicable ISO or ASME requirements for geometrical product specifications, CMM performance, calibration laboratories, uncertainty evaluation and conformity decisions. The laboratory must identify the governing standard rather than mixing incompatible definitions of datums, form error, filtering or acceptance limits.

Best practice includes:

• Define the measurand and acceptance rule before collecting data.
• Match sensor capability, resolution and uncertainty to tolerance.
• Stabilise temperature and document environmental conditions.
• Use traceable artefacts and maintain calibration records.
• Qualify probes and verify optical scale and distortion.
• Use functional datum alignment for acceptance unless another rule is specified.
• Report sampling density, filtering, fitting and registration methods.
• Investigate residuals and repeatability, not only average values.
• Preserve raw data and software configuration.
• Treat automation and AI as controlled measurement processes, not unquestioned authorities.

Chapter Summary

Optical, coordinate and non-contact metrology provide complementary ways to measure small profiles, discrete geometric features and complete free-form surfaces. Toolmakers’ microscopes and profile projectors are valuable for magnified two-dimensional inspection. Vision systems automate image-based feature extraction but depend strongly on lighting, optics and segmentation. CMMs provide versatile, traceable coordinate measurement through controlled probing, feature fitting and datum alignment. Laser and structured-light scanners collect dense three-dimensional data rapidly, although reflectivity, occlusion, calibration and registration affect their results.

A reliable inspection depends on the entire measurement chain: sensor calibration, workpiece preparation, probing strategy, coordinate transformation, fitting rule, environmental control, uncertainty evaluation and reporting. Functional datum alignment should not be confused with unconstrained best-fit alignment. Point clouds and colour maps are powerful diagnostic resources, but they require documented processing and independent verification. Contemporary automation and AI can improve productivity only when their outputs remain validated, traceable and appropriately governed.

Key Terms

Optical comparator: A profile projector that magnifies and displays a component silhouette.

Toolmakers’ microscope: A microscope with precision stages for measuring small features.

Machine vision: Image-based sensing and software analysis for inspection or measurement.

CMM: Coordinate measuring machine that determines three-dimensional feature coordinates.

Touch-trigger probe: A probe that records a point when contact is detected.

Scanning probe: A probe that continuously collects points along a surface path.

Probe qualification: Determination of effective probe geometry using a reference artefact.

Datum: A theoretically exact reference used to establish a coordinate system.

3–2–1 alignment: A datum-construction method using three, two and one constraints.

Best-fit alignment: Alignment that minimises a defined discrepancy between measured and nominal data.

Point cloud: A collection of measured spatial points.

Registration: Transformation of data sets into a common coordinate frame.

ICP: Iterative closest point algorithm for registering overlapping point sets.

Structured light: Three-dimensional measurement using projected coded patterns.

Triangulation: Depth determination from known geometric relationships between viewing and projection paths.

Least-squares fit: Fit that minimises the sum of squared residuals.

Uncertainty: Quantitative indication of doubt associated with a measurement result.

Traceability: Property linking a result to references through a documented calibration chain.

Residual: Difference between an observed point and a fitted or nominal feature.

Review Questions

Short Answer Questions

1. What is the principal difference between a toolmakers’ microscope and a profile projector?
2. Why is probe qualification required before CMM measurement?
3. Define datum alignment and best-fit alignment.
4. State two limitations of laser scanning.
5. What information should accompany a point-cloud colour map?

Long Answer Questions

1. Explain the operating principles, applications and limitations of toolmakers’ microscopes, profile projectors and machine-vision systems.
2. Describe a CMM measurement plan for a component containing a datum plane, bore and bolt-hole pattern.
3. Explain how structured-light scanning reconstructs a surface and discuss the effects of calibration, reflectivity and occlusion.
4. Compare least-squares fitting, minimum-zone evaluation and best-fit alignment in dimensional inspection.
5. Discuss calibration, verification, environmental control and uncertainty in an advanced metrology laboratory.

Analytical / Application Questions

1. A free-form casting has a glossy surface, deep recesses and a tight sealing-surface tolerance. Select a measurement strategy, justify the sensor and describe how you would control registration and uncertainty.
2. A part passes under a free best-fit alignment but fails when aligned to its functional datums. Analyse why this can occur and recommend the correct acceptance procedure.
3. Design an inspection plan combining a CMM and a scanner for a pump housing. Specify the features assigned to each system, alignment method, verification checks and information to retain in the report.

Multiple Choice Questions with Answers

Question 1

Which instrument is primarily intended for magnified measurement of small two-dimensional features?

A. Toolmakers’ microscope

B. Pressure transducer

C. Surface plate only

D. Thermocouple

Correct Answer: A. Toolmakers’ microscope

Question 2

In the magnification relationship [[M:I/O]], what does I represent?

A. Illumination intensity

B. Image size

C. Index of refraction

D. Inspection interval

Correct Answer: B. Image size

Question 3

What is the main purpose of probe qualification on a CMM?

A. To determine effective probe geometry

B. To increase workpiece hardness

C. To remove all datum requirements

D. To measure room humidity

Correct Answer: A. To determine effective probe geometry

Question 4

Which alignment is normally most appropriate for a functional acceptance decision?

A. Unconstrained visual alignment

B. Functional datum alignment

C. Random point alignment

D. Alignment to the largest colour-map deviation

Correct Answer: B. Functional datum alignment

Question 5

A laser scanner is especially useful for which task?

A. Measuring a free-form surface with dense data

B. Determining material chemical composition

C. Measuring electrical resistance

D. Hardening a machined surface

Correct Answer: A. Measuring a free-form surface with dense data

Question 6

What is a major risk when using ICP registration?

A. It can converge to an incorrect local solution

B. It always destroys all point data

C. It requires no initial alignment

D. It eliminates calibration requirements

Correct Answer: A. It can converge to an incorrect local solution

Question 7

For independent uncertainty components of 3 µm, 4 µm and 2 µm, the combined standard uncertainty is approximately:

A. 2.00 µm

B. 5.39 µm

C. 9.00 µm

D. 29.0 µm

Correct Answer: B. 5.39 µm

Question 8

Which factor commonly affects optical measurement of glossy metal?

A. Surface reflectivity

B. Nominal drawing scale only

C. Material density alone

D. The colour of the inspection room walls only

Correct Answer: A. Surface reflectivity

Question 9

What does a residual represent in geometric fitting?

A. The difference between an observed point and a fitted or nominal feature

B. The maximum machine travel

C. The probe storage temperature

D. The number of datum features

Correct Answer: A. The difference between an observed point and a fitted or nominal feature

Question 10

Why should a colour map not be treated as the complete measurement result?

A. It may hide alignment, filtering, scale and uncertainty information

B. It contains no visual information

C. It can only be used with contact probes

D. It automatically reports material composition

Correct Answer: A. It may hide alignment, filtering, scale and uncertainty information