By Lilian DAVIN, General Director of CODA SYSTEMS | July 2026
In an industrial context where quality, traceability, and productivity are non-negotiable requirements, industrial vision has established itself as an essential technology. Among its most critical applications, dimensional measurement holds a central place.
But while vision systems offer remarkable performance, their reliability depends on a set of best practices that too many manufacturers still overlook. At CODA Systèmes, we support our clients daily in the design and deployment of vision measurement solutions. Here’s what we have learned in the field.
A STRUCTURED APPROACH: THE 5M FRAMEWORK
Before delving into the details of good technical practices, it is useful to rely on a proven quality tool: the Ishikawa diagram, also known as the 5M diagram.
Initially developed for the analysis of defects and non-conformities in production, the 5Ms provide a universal framework that is particularly relevant for structuring the design and diagnosis of an industrial vision measurement system.
The 5Ms refer to the five major families of factors that influence the quality of a result:
M | Domain | Industrial vision application |
Measuring instrument | The measuring piece | Geometry, surface condition, material, reflectivity |
Environment | The environment | Ambient lighting, vibrations, temperature, dust |
Material | The equipment | Camera, optics, dedicated lighting, integration mechanics |
Method | The measurement process | Algorithms, parameters, calibration, procedures |
Labor | The operators | Training, skills, interpretation of results |
💡 The key idea: a failure in the accuracy or reliability of a vision measurement system always originates from one of these five areas. By systematically reviewing them (during both design and diagnosis), we ensure that nothing is overlooked.
Let's now structure our best practices around these 5Ms.
M1 / Measured quantity: thoroughly understand the piece to be measured
It's a classic mistake: rushing to get a sensor or a camera without properly defining the measurement need or analyzing the part itself.
The measurand (that is, the part to be inspected) is the starting point of any reflection. Its physical characteristics directly condition all the technological choices that will follow.
The questions to ask oneself:
- What is the quantity to be measured? (linear dimension, diameter, angle, flatness, roughness…)
- What is the acceptable tolerance?
- What is the production rate and the feed speed?
- What is the material? (shiny metal, matte plastic, clear glass, black rubber…)
- Is the piece reflective, transparent, textured, deformable?
- Is there natural variability in the part (color, brightness, slight geometric variations)?
What this implies...:
The answers to these questions directly determine the choice of technology: 2D camera, laser triangulation sensor, 3D profilometer, stereoscopic camera, or structured light. There is no universal solution.
A polished steel part is not treated the same as a loaded polyamide part. A black part on a black background will require a radically different lighting strategy than a white part.
💡 CODA Systèmes Advice: Always conduct a feasibility study with representative parts, including borderline and defective parts, before any investment. It's a significant time and money saver.
M2 / Environment: mastering the measurement environment
The environment is the setting in which the vision system must operate. It is often the most difficult factor to control, as it evolves over time and can be a source of insidious drift.
Environmental disruptors to identify:
- Ambient lighting is the primary threat. Variations in natural light (depending on the time of day, weather, seasons) or workshop lighting (flashing neon lights, reflections from nearby machines) are a major source of instability. Optical isolation or synchronization of the lighting with the acquisition is often necessary.
- The vibrations transmitted by nearby machines, conveyors, or presses can blur images and introduce measurement errors, especially at high resolution. Anti-vibration mounts or decoupled frames are then necessary.
- The temperature affects the expansion of mechanical structures, the parts themselves, and optics. In environments with high thermal gradients (near furnaces, welding areas, or outdoors), an uncompensated thermal drift can undermine a calibration that is otherwise rigorous.
- Cleanliness : dust, cutting oils, chips, or water splashes degrade optics and lighting. Waterproof enclosures (appropriate IP rating) and automatic blowers are often essential.
Best practices
- Conduct an environmental audit before any installation
- Plan for a housing or a cover to isolate the system from external disturbances
- Monitor and log ambient conditions to correlate potential measurement drifts with environmental events
💡 CODA Systèmes Advice : Never validate a system solely in laboratory conditions. Test it in real production conditions, at different times of the day and in different line configurations.
M3/ Equipment: choose and integrate the right tools
The hardware encompasses all the physical components of the system: camera, optics, dedicated lighting, additional sensors, and mechanical integration structure.
Dedicated lighting: the number one factor in measurement quality
If we had to focus on just one critical parameter, it would be lighting. Poor lighting undermines the performance of even the best optical system.
- Contrast what you want to measure : the lighting should highlight the contours or characteristics to be measured
- Choose the right type of lighting :
Coaxial lighting: ideal for flat surfaces and engravings
Backlighting: perfect for measuring contours and silhouettes
Raking light: reveals the sharp edges and textures
Dome lighting: uniform, recommended for reflective surfaces
- Stabilize the light source : prefer LED lighting with electronic regulation, which is much less sensitive to thermal drifts.
Optics and resolution
The system's resolution must be consistent with the required accuracy. A well-known empirical principle states that the physical resolution of the sensor (size of a pixel reported in object space, in µm/pixel) must be multiplied by a safety factor that depends on the type of lighting used:
- Coefficient 3 for backlighting : the sharp silhouette of the object against a bright background allows for very precise sub-pixel localization of the edges
- Coefficient 10 for direct lighting : the edges are less sharp, more noisy, which degrades detection accuracy
The obtained value (resolution × safety factor) must then be at least 10 times lower than the measurement tolerance.
However, it is also important to note these points of caution:
- Optical distortion : every lens introduces geometric distortion, especially at the periphery of the field. Calibration of the optical system is essential.
- The depth of field : out-of-focus parts induce measurement errors.
- Telecentric lenses : for metrology applications, they eliminate the perspective error related to distance variations and are often essential.
The mechanics of integration
A vision measurement system is only as good as its mechanical support.
- Securely mount the camera and lighting on a stable frame, ideally made of INVAR 36 or cast iron for high-precision applications.
- Mastering the positioning of parts: a vision system cannot compensate for poor positioning repeatability
- Protect the system from splashes, dust, and thermal shocks
- Precisely synchronize the acquisition with the movement of the parts in case of on-the-fly measurement
💡 CODA Systèmes Advice: Never skimp on the lighting and mechanical budget. This is often where 80% of a measurement system's performance is determined.
M4/ Méthode : définir des processus de mesure rigoureux
The most efficient hardware yields nothing without a solid method. The method encompasses image processing algorithms, calibration procedures, and decision rules.
Calibration: the foundation of all reliable measurement
A non-calibrated vision system does not measure, it estimates.
The key steps of a rigorous calibration:
- Geometric calibration: correction of distortion and the pixel/mm scale factor using a certified calibration target (checkerboard, point grid…)
- Calibration of the complete measurement chain: under real production conditions, with reference pieces traceable to national standards (COFRAC, ISO 10360…)
- MSA Verification (Measurement System Analysis): study of repeatability and reproducibility (R&R) to qualify the capability of the measurement system
- Documentation and traceability: each calibration must be recorded, dated, and archived
Calibration is not a one-time act. Plan for periodic re-calibration procedures, especially after any shock, change in thermal environment, or mechanical intervention.
Image processing algorithms
- Edge detection: the basis for many dimensional measurements
- Geometric fitting: circles, lines, ellipses, planes adjusted by least squares
- Shape correlation: robust localization of the object in the field of view
- Adaptive filtering and thresholding: to overcome residual lighting variations
- Strict definition of measurement areas (ROI – Region of Interest): a ROI that is too large includes noise, while a ROI that is too narrow loses information.
Statistical monitoring in production
A measurement system does not only exist at the time of its commissioning. Integrate from the design stage:
- The SPC (Statistical Process Control): monitoring measurement drifts over time
- Alert and action thresholds: distinct from product tolerances, to react before the system goes out of control
💡 CODA Systèmes Advice: Document your algorithmic parameters and procedures precisely. In case of product or line changes, you will quickly find and adapt your configuration.
M5/ Workforce: enhancing human skills
It is the M that is most often forgotten in industrial vision projects, as we tend to think that automation frees itself from the human factor. This is a mistake.
The workforce (operators, technicians, quality engineers) remains at the heart of the performance of a measurement system for three essential reasons.
1. The configuration and the settings
A vision system is configured by humans. Poorly chosen parameters (detection thresholds, measurement areas, decision rules) can lead to unacceptable rates of false positives or false negatives, even with the best equipment in the world.
2. Monitoring and Interpretation
A trained operator is capable of detecting an abnormal drift, interpreting an alarm, and distinguishing a real defect from an image artifact. Without training, these signals are ignored or, worse, lead to poor decisions.
3. First-level maintenance
Regular cleaning of optics and lighting, visual inspection of mechanical condition, initiation of periodic verification procedures: all simple yet crucial tasks that fall under the field workforce.
Best HR and organizational practices:
- Train the operators during the commissioning and at each system upgrade
- Write clear procedures for monitoring, verification, and first-level maintenance
- Involve the quality teams from the system design phase, not just during the acceptance
- Create a measurement culture : raise awareness among teams about the importance of metrological reliability and the consequences of a poorly maintained system
- Clearly define the responsibilities : who calibrates? Who alerts? Who decides to stop the line ?
💡 CODA Systèmes Advice: During our integration projects, we systematically offer training sessions tailored to each profile: operators, maintenance technicians, and quality engineers. A well-understood system is a well-used system.
SYNTHESIS : THE 5M AS A PERMANENT FRAMEWORK FOR ANALYSIS
The 5Ms are not only used to design a vision measurement system. They provide a permanent framework for diagnosing any drift or failure:
Unreliable measurement system?
- Measuring device → Has the part changed? New batch, new supplier?
- Environment → Has the environment evolved? Season, workshop rearrangement?
- Equipment → Dirty optics? Diminished lighting? Mechanical play appeared?
- Method → Outdated calibration? Modified parameters? Procedure not followed?
- Labor → New operator? Insufficient training? Handling error?
By systematically going through these five branches, one quickly identifies the root cause and provides a targeted response rather than navigating blindly.
CONCLUSION
Industrial vision measurement is a powerful but demanding technology. Its reliability does not solely depend on the quality of the hardware or software: it is the result of a rigorous and comprehensive approach, which the 5M framework allows to be structured effectively: from the design phase to daily monitoring in production.
Measured quantity, Environment, Material, Method, Labor: these five dimensions must be treated with the same rigor. Neglecting any one of them is to accept a fragility in one's measurement system.
At CODA Systems, we leverage our expertise to serve manufacturers by designing robust, precise vision measurement solutions tailored to your production constraints. From feasibility studies to turnkey integration, including training for your teams, we support you at every step.
CODA SYSTEMS LLC – Experts in industrial vision systems www.coda-systemes.fr