Machine Vision Selection | Choosing the Right Band = Half the Battle!

‍‍‍‍‍‍In machine vision projects, choosing the wrong camera band renders lighting, lenses, and algorithms useless.

Many engineers start with visible light and white light illumination, only to fail: transparent parts are unclear, internal defects are undetectable, fluorescent flaws cannot be measured, and high-temperature scenarios are unworkable.

This guide explains how to select from UV (Ultraviolet), VIS (Visible), NIR (Near-Infrared), SWIR (Short-Wave Infrared), MWIR/LWIR (Mid/LongWave Infrared), including suitable applications and key pitfalls. After reading, you can select directly for implementation.

Ⅰ. First Understand: The Band Determines “What Can Be Seen”

Industrial imaging does not rely only on light visible to the human eye. Different wavelengths correspond to four capabilities:

penetration, fluorescence, scattering, and thermal radiation,

which directly determine whether defects can be detected.

One-sentence logic:

• Detect micro-surface defects / fluorescence → Choose UV

• Detect color / appearance / dimensions → Choose VIS

• Detect plastic penetration / color interference elimination → Choose NIR

• Detect silicon wafers / moisture / deep penetration → Choose SWIR

• Detect temperature measurement / thermal defects / total darkness → Choose IR

Ⅱ. Complete Analysis of 5 Bands: Features + Applications + Pitfalls

1. UV Band (200–400nm)

Core Capabilities: Fluorescence excitation, strong scattering, detection of tiny surface defects

• Common: UVA 365nm (most widely used)

• Advantages:

  • Clearly reveals micro-scratches, cracks, smudges, fingerprints on glass, wafers, and cover plates

  • Fluorescence imaging for glues, inks, oils, and anti-counterfeit labels

  • Maximizes contrast for micro-defects in transparent materials

  
  

• Limitations: Requires dedicated UV lenses + UV light sources; higher cost; protective measures needed

• Must-use scenarios: Semiconductor cracks, phone screen scratches, PCB glue residue, fluorescence anti-counterfeiting, cleanliness inspection

  

  MindVision UV Industrial Camera Spectral Graph
  

2. VIS Band (400–700nm)

Core Capabilities: Universal, low-cost, intuitive debugging

• The most mainstream in industry; widest selection of lenses/light sources; best cost-performance

• Preferred monochromatic light:

  • Red: Metal / PCB / OCR, reduces reflection

  • Green: Precision measurement, high resolution

  • Blue: Plastics / printing / low-reflection surfaces

  
  

• Must-use scenarios: Appearance inspection, dimension measurement, color recognition, barcode/OCR, assembly positioning, general defect detection

  

  MindVision Visible-Light Industrial Camera Spectral Response
  

3. NIR Band (700–1400nm, common 850/940nm)

Core Capabilities: Penetrate non-metals, eliminate color interference

• CMOS-compatible; NIR-enhanced models available at controllable cost

• Advantages:

  • Penetrates plastics, paper, silica gel, fog/haze

  • Ignores color, focuses only on structure/internal features

  • 940nm is invisible to humans, no glare

  
  

• Must-use scenarios: Bottle liquid level, internal paper wrinkles, silicon/photovoltaic cracks, food moisture/foreign objects, internal electronic components

  

  MindVision NIR Band Industrial Camera Spectral Graph
  

4. SWIR Band (900–1700nm)

Core Capabilities: Industrial “X-ray vision”, more penetrating and specialized than NIR

• InGaAs sensors; unique absorption/transmission for silicon, water, and plastics

• Advantages:

  • Silicon is transparent above 1100nm → Internal defects in wafers/LEDs

  • Strong water absorption at 1450nm → Accurate moisture/rot detection

  • Penetrates PE/HDPE and smoke

  
  

• Must-use scenarios: Semiconductor wafer inspection, Mini/Micro LED, food rot/moisture, high-temperature metals, smoke-penetrating observation

MindVision SWIR Band Industrial Camera Spectral Graph

5. MWIR/LWIR Band (3–5μm / 8–14μm)

Core Capabilities: No light source needed; measures temperature and thermal distribution

• Passively receives thermal radiation, usable in complete darkness

• Advantages:

  • Non-contact temperature measurement, abnormal temperature difference, thermal failure

  • Stable imaging in rain, snow, fog

  
  

• Limitations: Low resolution, high cost (cooled types are more expensive)

• Must-use scenarios: Welding temperature measurement, battery thermal runaway, metallurgical high temperatures, predictive maintenance, total darkness monitoring

  

MindVision SWIR Band Industrial Camera Spectral Graph

Ⅲ. Direct Application: 3-Step Selection (No Pitfalls)

Step 1: Ask Yourself 3 Questions

1. Need color / appearance? → Visible light

2. Need micro-surface defects / fluorescence? → UV

3. Need penetration / internal / temperature measurement? → NIR/SWIR/IR

Step 2: Select Directly by Material

• Glass / cover plate / wafer → UV + SWIR

• Plastics / packaging / liquid level → NIR + SWIR

• Metal / PCB / OCR → Visible light (Red/Green)

• Food / agricultural products → NIR + SWIR (moisture/defects)

• High temperature / welding / thermal failure → LWIR/MWIR

Step 3: Matching Accessories Are Essential

• UV: Dedicated UV lens + UV light source

• NIR/SWIR: IR-cut removal, anti-reflection coating

• Thermal IR: Special lenses, no illumination needed

• Monochromatic light: Use matching band-pass filters to reduce noise and boost contrast

Ⅳ. Collectible Selection Mnemonic (For Sharing)

Micro flaws & fluorescence: choose UV

Appearance & size: choose VIS

Penetration & de-coloring: choose NIR

Silicon & moisture see-through: SWIR

Temp & dark: thermal IR

Right band = high efficiency

Ⅴ. Final Reminders: Avoid 3 Misconceptions

1. More expensive = better: Visible light is sufficient for general inspection. High-end bands are for “necessity”, not “standard”.

2. Only care about the camera, not the lens: Mismatched band and lens = useless imaging.

3. Random light source matching: Using white light for UV, visible light for NIR — completely ineffective.

Summary

Industrial camera band selection is essentially using the right light capabilities:

• See defects invisible to the human eye

• Penetrate materials ordinary light cannot

• Measure temperature and compositional differences

Choosing the right band increases project success rate by 80%, halves debugging time, and greatly reduces algorithm difficulty.

You may contact us at chenguo@mindvision.com.cn to gain more in-depth technical insights and practical applications in the fields of machine vision and optical imaging.