Inline vs Offline Optical Measurement: How to Choose the Right Inspection System

By the POMEAS Metrology Applications Team · Reviewed by Tiger Xie · Published 13 October 2026

Article ID: POMEAS-KN-INLINE-VS-OFFLINE-EN-v1 Author: POMEAS Metrology Applications Team · Reviewer: Tiger Xie · Date: 2026-10-13

1. The question is not "which machine is better"

Most teams ask us the wrong version of this question. They ask "should we buy an inline system or a measuring machine?" — as if the two were competing at the same job.

They are not. Inline and offline optical measurement sit at different points in the production flow, and they answer different questions.

• Offline measurement answers: "Is this part good, and can I prove it?" You take a part (or a sample) to a measuring station, place it on a stage, and measure it in full detail. Speed matters less than completeness, traceability and the ability to measure features that are hard to reach.
• Inline measurement answers: "Is the process still producing good parts — right now?" The measurement happens on, above or beside the production line, on every part or on a fixed sampling interval, without removing the part from the flow.

A shop that needs 3 parts per shift inspected in full detail and a shop that needs 20,000 parts per shift checked on 4 critical dimensions are not making the same decision. They should not buy the same architecture.

This article gives you the six factors that actually decide it, with the numbers POMEAS publishes for its own systems so you can run the comparison against your own parts.

2. The six factors that decide it

Notice that "accuracy" is deliberately not #1. That is the most common mis-selection we see — see §4.

3. What POMEAS actually builds for each job

3.1 Offline: the instant vision measuring system

The POMEAS IMAGE 3 Instant Vision Measuring System is a benchtop optical measuring system for one-click dimensional inspection. It uses optical imaging with telecentric lens technology, a high-resolution CMOS camera and AI-Image measurement software to measure length, width, diameter, radius, angle, distance, position and profile features, and to output a QC result.

Its role in the flow is a measurement station: a part is placed on the stage, the system images it, extracts the features and produces the dimensional report. (For how to pick a model, see the IMAGE 3 selection guide.)

3.2 Inline: the telecentric measurement system

The POMEAS HM Telecentric Measurement System is built for the opposite constraint. It measures non-contact, in-line, on the production line, and does not require the process to stop. Its architecture is:

• Dual-side telecentric optical system with CMOS shadow profile imaging — the part is measured as a silhouette, which is why edges stay stable when the part's surface finish changes.
• InGaN green LED parallel illumination — collimated light, so an edge does not move when reflectivity changes.
• IP64 measurement head, 0–45 °C, ≤85 % RH — it is designed to live on a shop floor, not in a metrology lab.

The HM series is offered in four measuring ranges:

Read this table carefully — it contains the single most common specification mistake in this category. The ±0.05 μm figure belongs to the HM-1015 and is a repeatability value, not a measurement accuracy, and it does not apply to the other three models. Accuracy, repeatability and the usable high-precision area all change with the model. POMEAS states this explicitly on the product page, and we repeat it here because the number is routinely quoted out of context in buyer comparisons.

System-level HM specifications:

3.3 The third option most buyers miss

The two systems above are not mutually exclusive, and in our experience the most stable production quality systems use both: an inline system on the line for the 3–6 dimensions that actually drift, and an offline measuring system in the QC room for full-feature verification and for the audit trail that the inline system is not designed to produce.

If you are choosing between the two for a single budget line, §5 and §6 give the decision rules.

4. Accuracy, repeatability and why the numbers are not directly comparable

This is where most comparisons go wrong.

• Repeatability describes how tightly the same system repeats when measuring the same part in the same position. It is a stability figure.
• Accuracy describes how close the reading is to the true value. It includes the calibration chain, the reference artefact, fixturing, temperature and operator technique.

An inline system that repeats to ±0.05 μm is not automatically "more accurate" than a benchtop measuring machine. It means that, under fixed conditions, its readings scatter less. Whether that translates into accuracy depends on how the system is calibrated and how the part is presented to it.

Two practical consequences:

1. Never specify inline repeatability as the required measurement accuracy in a purchase specification. They are different claims. Convert them properly — see the calibration and measurement stability FAQ and object-side resolution vs measurement accuracy, which walk through the chain.
2. Test the method, not just the machine. POMEAS publishes TR-001, Machine Vision Optical Imaging Systems — Performance Test Methods: 16 test methods covering FOV, working distance, MTF resolution, TV and geometric distortion, depth of field, telecentricity and zoom parfocality, with calculation methods, acceptance criteria and a report template. If a supplier cannot tell you which of those methods was used to produce a number, the number is not yet usable.

5. Cycle time: the factor that usually decides it

Cycle time is where inline and offline separate decisively, and it is calculable before you buy anything.

Worked example — a shaft with 4 critical dimensions, one shift of 7.5 productive hours

The offline figure is arithmetic on the handling and measurement steps. The inline figure is not a throughput claim — an inline head's exposure is short (35–150 μs), but real throughput is set by the line takt and the trigger/handshake design, not by the camera. POMEAS lists full-field acquisition at ~30 fps and local 1/8-field at up to ~200 fps for the HM series; those are acquisition figures and must be multiplied by your actual part-present time and reject-handling logic.

The honest comparison is therefore not "20 s vs 0.1 s". It is:

Offline throughput = available operator hours ÷ (handling + measurement) → roughly 1,000–1,500 parts per station per shift in the example above.
Inline throughput = line takt → whatever the line already produces, because the measurement overlaps the process.

If your requirement is sampling (say 30 parts per shift), offline wins on cost, flexibility and measurement scope. If your requirement is 100 % inspection at line speed, offline cannot get there at any reasonable headcount, and the decision is made for you.

6. When offline is the right answer

Choose an offline measuring system such as the IMAGE 3 when:

• You need many features and full GD&T on a part, including features that are shadowed from a single in-line viewpoint.
• The part is expensive or low-volume, and the cost of a fixture that survives line conditions is not justified.
• You need a traceable report per batch and an audit trail, rather than per-unit data.
• The measurement itself is the bottleneck you are fixing — replacing manual callipers and micrometers with an optical station. (This is the most common first step we see: manual → offline optical → then, later, inline.)
• You are still learning which dimensions actually drift. Measuring offline first tells you what to monitor inline. This is the sequence we recommend to most customers.

7. When inline is the right answer

Choose an inline system such as the HM when:

• 100 % inspection is required by the customer, the standard or the risk of a single escape — medical implants and fasteners are the classic cases, where the cost of one bad part leaving the plant dwarfs the system cost.
• The line cannot stop and parts cannot be carried to a station.
• Re-fixturing changes the measurement — if the dimension you care about only exists in the part's in-process state, an offline measurement measures a different thing.
• You already know the 3–6 dimensions that drift and want them monitored continuously, with alarms.
• Per-unit traceability is required, and a paper batch report is not enough.

8. Decision checklist

Answer these before requesting a quotation. It takes ten minutes and removes most of the risk:

1. Which dimensions are critical (not "all dimensions")? List at most six.
2. What is the tolerance on each, in μm?
3. What is the required sampling rate — AQL, fixed interval, or 100 %?
4. What is the line takt, in seconds per part?
5. Can the part be removed from the flow? Does removal change the measurement?
6. What is the part envelope (max diameter/width × height) and its surface finish (polished, matte, transparent, black)?
7. Is per-unit data traceability required, and in what format?
8. Who handles a reject — operator, PLC, or robot?
9. What is the ambient environment (temperature range, coolant, dust, vibration)?
10. Is the goal replacing manual measurement (→ offline first) or catching escapes on a running line (→ inline)?

9. Frequently asked questions

Can an inline system replace a measuring machine entirely? Usually not, and it is not designed to. An inline head is optimised for a small, fixed set of dimensions at line speed. Full-feature verification, complex datums and the audit trail are still better served by an offline measuring system. Most mature quality systems run both.

Is ±0.05 μm the accuracy of the HM system? No. ±0.05 μm is the repeatability of the HM-1015 specifically. Repeatability is not accuracy, and the figure does not apply to HM-1040 (±0.08 μm), HM-1065 (±0.1 μm) or HM-1120 (±0.2 μm). Accuracy, repeatability and the high-precision area all vary by model, and positional accuracy within the high-precision area is specified separately (up to ±0.3 μm).

How fast is inline measurement, really? Exposure time is 35–150 μs depending on model, and acquisition runs at approximately 30 fps full-field or up to approximately 200 fps in local 1/8-field mode. Your actual part rate is set by the line takt, the trigger arrangement and the reject-handling logic — not by the camera specification alone.

Can the HM system measure transparent or highly polished parts? The HM uses a dual-side telecentric optical system with green LED parallel illumination and CMOS shadow imaging. Because the measurement is based on the shadow outline rather than surface reflection, it is far less sensitive to reflectivity than a reflected-light system. Transparent parts need to be evaluated on samples — please send parts for a test.

What about parts with internal features? Shadow-based inline measurement sees the silhouette. Internal bores, undercuts and hidden features generally need a different technique — a telecentric measurement station or a multi-sensor offline system. See Telecentric Lens for Dimensional Measurement and Inspection.

Can we start inline and add offline later? Yes, and that is often correct. But the reverse — starting offline and planning to "upgrade to inline" — requires that the critical dimensions are already identified, which is exactly what the offline phase is for.

Do we need the same fixture for both? No, and they generally should not share one. The inline fixture is a line fixture (hardened, repeatable, tolerant of coolant and swarf); the offline fixture is a metrology fixture (clean, adjustable, often with fine positioning). Forcing one fixture to do both compromises both.

10. Related POMEAS resources

• POMEAS HM Telecentric Measurement System for In-line Inspection
• POMEAS IMAGE 3 Instant Vision Measuring System for QC Inspection
• IMAGE 3 Selection Guide: Choose the Right Image Measuring Machine
• What Is a Flash Measuring Machine? One-Click Image Dimension Measurement
• Telecentric Lens for Dimensional Measurement and Inspection
• Migration case studies (offline → inline):
• Watch Case Dimension Measurement: Minutes to Seconds per Part
• Camera Module Die-Cut Measurement: 60 s to Under 5 s per Part
• Medical Bone Screw Dimensional Inspection Solution
• Automated Medical Tube Inspection System for Transparent Tubing
• Calibration and Measurement Stability FAQ
• TR-001 Machine Vision Optical Imaging Systems — Performance Test Methods

11. Talk to an applications engineer

If you send us the answers to the checklist in §8 plus a drawing of the part, we will tell you which architecture fits — including when the answer is "you do not need an inline system yet".

• Email: [email protected]
• WhatsApp: +86 185 9810 2007

Data integrity note

• All POMEAS specifications in this article are taken from the live product pages for the HM and IMAGE 3 systems and from POMEAS TR-001. No parameter has been estimated.
• The cycle-time example in §5 is arithmetic on stated assumptions (12 s handling, 8 s measurement, 7.5 productive hours). It is an illustration, not a measured production result.
• No search volume, ranking or keyword-difficulty figures are given, because POMEAS has no verified source for them. Where a number would have to be guessed, it is marked as to-be-confirmed rather than invented.
• Application examples are drawn from POMEAS customer case files.