Industrial vision inspection systems are at the heart of modern automated production. Their hardware selection directly determines inspection accuracy, stability, and return on investment. The performance of a vision system depends on its weakest link – choosing the wrong camera, lighting, or lens will cripple even the best algorithm in the field. The global machine vision market reached $16.5 billion in 2025 and is projected to grow to $42 billion by 2035 at a 9.8% CAGR – making proper hardware selection mission-critical for system integrators and equipment manufacturers. This guide systematically breaks down the selection methodology for cameras, lenses, and lighting in industrial vision projects, covering the full workflow from requirement analysis to collaborative matching.

1. Preparation: Translating Requirements into Parameters

Every vision project starts with accurately understanding the inspection needs. Below we analyse the five essential preparatory steps.

Step 1: Define the inspection objective.

Describe in concrete terms what you need to "see" – missing parts, barcode reading, solder joint inspection, gap measurement, or robot guidance? Different objectives impose completely different hardware demands.

Step 2: Quantify key parameters.

Field of View (FOV): How large an area must the camera capture at once? For example, a single chip (10mm×10mm) or a whole PCB (300mm×200mm)? Typically the FOV should be 10–20% larger than the object for positioning tolerance. Working Distance (WD): The distance from the lens front to the object surface – constrained by mechanical structure, must be determined early. Inspection accuracy: What is the smallest feature size to detect? Accuracy = FOV (one side) / camera resolution (pixels in that direction). Example: FOV 100mm wide, 2000 pixels → 0.05mm per pixel. Line speed: Is the object stationary or moving? Speed dictates required frame rate and exposure time. Environmental conditions: Vibration, temperature, dust, space constraints – determine protection ratings and mounting.

Step 3: Prepare test samples.

Good samples are not enough – the test set must include real defects, acceptable variations, dirty surfaces, reflective parts, different batches, and slightly misaligned positions.

2. Camera Selection: Sensor Choice Dictates Image Quality

The camera converts light into a digital image. Selection must balance resolution, frame rate, sensor type, and interface protocols.

2.1 Resolution and Detection Accuracy

Resolution determines the smallest detail you can see. Selection principle: allocate at least 3–5 pixels across the smallest defect.

Resolution calculation example: Inspecting solder bridging defects as narrow as 0.1mm on a PCB with a 60mm FOV. Pixel pitch required = 0.1mm / 3 ≈ 0.033mm/pixel. Horizontal pixels = 60 / 0.033 ≈ 1818. Thus a minimum 2MP (1920×1200) camera is required, but 5MP (2592×1944) is recommended for headroom.

2.2 Frame Rate and Motion Matching

Frame rate determines how many images per second can be captured, thus the production line throughput. Key considerations:

Motion blur calculation: If an object moves more than 1 pixel during exposure, image blur occurs. Example: line speed 1m/s, exposure 10ms → movement = 1m/s × 0.01s = 10mm. If pixel pitch is 0.05mm, that's 200 pixels of blur – image completely useless.

2.3 Sensor Type: Global Shutter vs Rolling Shutter

This is one of the most critical decisions in camera selection.

Industry rule: Any scenario involving motion, vibration, or robot guidance must use global shutter. Only if the object is perfectly stationary and lighting is sufficient can rolling shutter be considered as a cost-saving option.

2.4 Pixel Size and Sensitivity

For the same resolution, pixel size (physical dimensions) directly affects image quality:

• Large pixel (e.g., 5.5μm): larger full-well capacity, wider dynamic range, lower noise in low light, but larger sensor diagonal → higher lens cost
• Small pixel (e.g., 2.2μm): higher resolution possible in a smaller package, but diffraction-limited (lens resolution bottleneck), more noise in low light

Rule of thumb: all else equal, prefer cameras with pixel size ≥ 4μm.

2.5 CCD vs CMOS: The Technology Evolution

• CCD (Charge-Coupled Device): global shutter is standard, excellent image quality (low noise, high dynamic range), but high power consumption, slow readout, expensive. Best for high-precision static inspection.
• CMOS (Complementary Metal-Oxide-Semiconductor): now dominant in industrial vision – fast, low power, and modern CMOS supports global shutter (e.g., Sony Pregius series). High-end CMOS rivals CCD in image quality at a better price point.

Conclusion: Over 90% of industrial vision applications use global-shutter CMOS. CCD remains only for ultra-low-noise scientific measurement.

2.6 Interface Selection

Recommendation: Most projects use GigE; for high-res/high-frame-rate, use 10GigE or USB3.0.

3. Lens Selection: Core of Optical Path Design

The lens projects the object image onto the camera sensor. A wrong lens introduces distortion, insufficient light, or failure to resolve fine details.

3.1 Mount Compatibility: Start with the Mount

Lens mount must match camera mount. Common mounts:

Note: C-mount lenses can be used on CS-mount cameras with a 5mm adapter ring, but CS-mount lenses cannot be mounted on C-mount cameras (they will hit the sensor).

3.2 Image Circle Matching: Cannot Be Smaller than Sensor

Core principle: lens image circle diameter ≥ camera sensor diagonal. Otherwise vignetting or severe edge quality degradation occurs.

3.3 Focal Length: Determines FOV and Working Distance

Focal length formula: f = (WD × sensor dimension) / FOV dimension

Example: Sensor width 6.8mm (1/1.8"), FOV width 60mm, WD = 200mm. f = (200 × 6.8) / 60 ≈ 22.7mm → choose a 25mm fixed focal lens.

Fixed focal lenses are preferred in industrial vision – they offer the highest image quality and best distortion control. Zoom lenses sacrifice optical quality for flexibility and are not recommended for measurement applications.

3.4 Aperture and Depth of Field

• Large aperture (small f/#, e.g., f/1.4): more light, good for low light, but shallow depth of field
• Small aperture (large f/#, e.g., f/8): deeper DOF, good for objects with height variation, but requires stronger illumination

DOF formula (approximate): DOF ≈ 2 × permissible circle of confusion × f/# × (1 + magnification) / magnification² where magnification = sensor size / FOV size. Small magnification (large FOV) → deep DOF; large magnification (small FOV) → shallow DOF.

Example: If required DOF is 2mm but object height variation is 5mm, parts will be out of focus. You must close the aperture (e.g., f/2.8 → f/5.6) to increase DOF, and simultaneously increase illumination.

3.5 Resolution Matching (Nyquist Limit)

The lens's resolving power is described by its MTF (Modulation Transfer Function). To fully utilise the camera: Lens cutoff frequency ≥ camera Nyquist frequency = 1/(2×pixel size)

• Pixel 5μm → Nyquist = 100 lp/mm → lens MTF at 100 lp/mm should be > 30%
• Pixel 3.45μm → Nyquist = 145 lp/mm → higher-resolution lens required

Rule of thumb: Lens object-side resolution = 2×pixel size / magnification gives perfect matching.

3.6 Distortion Control

Distortion is barrel (wide-angle) or pincushion (telephoto). For measurement applications, distortion must be < 0.1%:

• Telecentric lenses: inherently low distortion (< 0.05%) – first choice for measurement
• Standard industrial lenses: 0.5%–2% distortion, correctable by software

Correction warning: software correction loses some FOV and resolution.

3.7 Special Lens Types

4. Lighting Selection: The Most Overlooked Yet Most Critical Element

You can have the most expensive camera and lens, but if the lighting is poor, the image will always be poor. The goal of lighting is not to "illuminate" but to make the target feature stand out clearly against the background.

4.1 Objectives of Good Lighting

• Maximise contrast: grey-level difference between feature and background ≥ 50 grey levels
• Minimise irrelevant information: suppress ambient light, reflections, and texture clutter
• Uniformity: illumination uniformity > 80% (brightness variation across FOV < 20%)

4.2 Common Lighting Types and Applications

Industry rule: About 80% of applications can be solved with ring lights, bar lights, or backlights.

4.3 Colour Selection (Light–Object Interaction)

Choosing the right colour dramatically improves contrast and reduces algorithmic burden.

Complementary colour principle:

• Object colour = reflected wavelength
• Same-colour light → strong reflection → bright image
• Complementary colour light → object absorbs that light → dark image → maximum contrast with brightly lit background

4.4 Brightness and Strobing Control

Strobing principle: Provide a high-intensity light pulse only during camera exposure; off otherwise.

Key requirement: Frame rate, exposure time, and strobe pulse width must satisfy: pulse width ≥ exposure time (slightly longer), and light recovery time < frame interval.

4.5 Lighting Controller Selection

Key feature: trigger synchronisation – external sensor signal → after programmed delay (to align with object arriving at FOV centre) → triggers camera and strobe.

4.6 Lighting Mounting Pitfalls

• Light must not block the camera view nor interfere with robot arms
• Distance from light to object must be moderate – too close → local overexposure; too far → light loss
• High angle (>45°) for smooth surfaces; low angle (<20°) for scratches
• Transparent/semi-transparent objects → backlight is the first choice
• Highly reflective surfaces → coaxial or dome light to eliminate specular reflection
• Outdoor or bright environments → add a shroud or use narrow-band filters

5. Collaborative Matching of the Three Elements

The camera, lens, and lighting are not selected independently – they form a single imaging chain. Any weakness in any link becomes the system bottleneck.

5.1 Resolution Chain Consistency

The system's final resolution is determined by "lighting contrast × lens MTF × camera MTF".

• High camera resolution but poor lens MTF → high-frequency detail lost → camera upgrade useless
• Low lighting contrast → defect invisible → no camera or lens can help

Correct workflow: select lighting first (ensure feature visibility) → select lens (ensure detail transfer) → select camera (ensure digital capture)

5.2 Spectral Matching

• Light wavelength must lie within the camera's spectral response range
• Standard CMOS: visible (400–700nm)
• UV cameras: UV light source + quartz lens (ordinary glass absorbs UV)
• SWIR (InGaAs): SWIR light source + specially coated lens
• Multi-spectral lighting: use apochromatic lenses to correct focus shift

5.3 Timing Synchronisation

• Camera exposure start, light pulse, and object position (encoder/sensor signal) must be precisely synchronised
• Common trigger delay function: external sensor sends signal → after programmed delay (to match object reaching FOV centre) → triggers camera and light
• Relationship: frame rate, exposure time, and light pulse width must satisfy: pulse width ≥ exposure time (slightly longer), and light recovery time < frame interval

5.4 Mechanical Fit

• Lens length + camera thickness + light height + mounting bracket = total installation space
• Draw a mechanical layout before selecting components to ensure they fit
• For space-constrained environments, use 90° prisms or mirrors to bend the optical path
• Allow for cable bend radius and connector clearance

5.5 Environmental Robustness

6. Quick Selection Reference Table

7. Common Selection Pitfalls and How to Avoid Them

8. Frequently Asked Questions

9. Summary: The Three-Step Selection Process

There is no "one-size-fits-all" formula, but there is a scientific methodology:

Step 1: Define requirements – accuracy, speed, environment, space – all in numbers.

Step 2: Bottom-up preliminary selection – choose resolution based on accuracy, frame rate based on speed, light type based on surface properties.

Step 3: Top-down validation – test the selected hardware with real samples. If any requirement fails, upgrade the bottleneck component until all specifications are met.

Three core principles:

• Define requirements first, then select hardware – derive parameters from the inspection target, not from a catalogue.
• The three elements must work together – camera, lens, and lighting form an "iron triangle"; any weak link becomes the system bottleneck.
• Test, don't guess – verify with real samples under real conditions. Lab perfection does not equal field reliability.

About selection support

Correct selection is critical to project success. Wrong choices can lead to insufficient accuracy, system instability, cycle-time failure, and ultimately production losses.

For application-specific hardware recommendations, our engineering team provides full support: requirement analysis, sample testing, solution design, and on-site commissioning. Whether you need a single camera or a complete vision system, contact us for professional selection guidance.