RGB NIR Camera: What It Is and How It Sees Near-Infrared
An RGB NIR camera records visible color (red, green and blue) and near-infrared light, roughly 700 to 1000 nm, from the same scene, usually through a single sensor and lens. You get a normal color image when the light is good and a usable infrared image when it is not, with no second camera and no mechanical filter swap. Sellers also call it an RGB-NIR multispectral camera, an RGB-IR camera or a color NIR camera. The idea is the same, but the hardware behind each name differs.
This guide covers why the combination matters, how one sensor separates two kinds of light, where these cameras earn their keep, and how to choose one. It also covers the tradeoffs that product pages tend to skip, such as color distortion in sunlight and lost resolution.
Vadzo designs board-level USB, MIPI CSI-2 and GigE camera modules for OEMs. Many projects in the markets we serve, such as surveillance, robotics and machine vision, need both color and NIR. So the advice here is practical: what to check, what to ask a supplier, and which routes to a color and NIR product exist.
Why an RGB-NIR camera matters
An RGB-NIR camera matters because visible light alone is an unreliable signal. It vanishes at night and shifts with every lamp, cloud and window. Near-infrared can be supplied by an illuminator your eyes cannot see. Putting both bands on one device gives you dependable images around the clock without doubling your hardware.
Where visible-only cameras fall short
Silicon image sensors respond to visible light from about 400 to 700 nm, and they keep responding into the near-infrared, up to about 1000 nm. A standard color camera discards that extra range on purpose. The red, green and blue filters also pass NIR, so designers place an IR-cut filter in front of the sensor to keep colors accurate. The price is a camera that goes blind when the light drops.
NIR imaging between 700 and 1000 nm behaves differently. It gives stronger structural contrast and stable imaging in variable lighting. That is why machine vision teams use it where ambient light cannot be controlled.
A camera built only for visible light stops working when the lights do.
What you gain with both bands in one camera
The benefit is not only night vision. You keep color for the cases that need it and add NIR for the cases that color cannot handle.
One optical path. One lens, one board and one cable replace a pair of cameras.
Color on demand. People and many AI models need color to read labels, clothing or status lights.
Invisible lighting. A 940 nm illuminator lights the scene for the sensor without lighting it for people.
Lower cost and size. You skip the second camera and the motorized filter.
How an RGB-NIR camera captures both on one sensor
A single-sensor RGB-NIR camera adds dedicated infrared pixels to the color filter array, removes the IR-cut filter, and uses image processing to split the visible and NIR signals. The sensor never "switches" modes. It captures both bands in every frame and the processor decides what to do with them.
The filter array and the missing IR-cut filter
A normal sensor uses a Bayer pattern of red, green and blue filters. An RGB-IR sensor swaps some of those positions for pixels that respond to infrared. Because the IR-cut filter is gone, the NIR light reaches the sensor and the IR pixels record it.
This costs resolution. Say a quarter of the pixels on a 5 MP sensor are IR. That leaves about 1.25 million IR samples and 3.75 million for color. The camera reconstructs a full-resolution image of each band by interpolation, so fine detail is softer than on a pure color or pure NIR sensor.
Separating visible and NIR signals
The hard part is NIR leakage. The red, green and blue pixels also respond to NIR, so under bright sunlight or incandescent bulbs, which emit a lot of infrared, colors turn strange. Fixing this takes ISP tuning that subtracts the infrared contribution from each color channel. Sony's IMX775 RGB-IR sensor does this on the chip, at the 940 nm wavelength used in vehicle cabins. Cheaper modules often rely on a generic pipeline, and you see the difference in a sunlit window.
The sensor is only half of an RGB-NIR camera. The color-correction tuning is the other half.
Four ways to build one, compared
The terms RGB-IR camera, color NIR camera and RGB-NIR multispectral camera cover several designs. Each one trades cost, resolution and simplicity differently.
Design | How it works | Strength | Tradeoff |
|---|---|---|---|
Single RGB-IR sensor | IR pixels sit in the filter array, no IR-cut filter | One sensor, both bands in every frame | Lower resolution per band, color contamination in sunlight |
Color sensor with switchable IR-cut filter | Filter in by day, out at night | Full resolution in each mode | Moving part, bands not captured together, night image is gray |
Two cameras, color plus NIR | Separate sensors, aligned in software | Best quality in each band | Higher cost, bigger size, calibration work |
Snapshot multispectral | Mosaic of band filters, often 4 channels | Measurement-grade simultaneous bands | Few pixels per band, such as 512 x 512 on a 1-inch sensor |
One more term causes confusion. An RGB-D camera adds depth, not infrared color. It often uses NIR light to measure that depth, but it is a different product.
What RGB-NIR cameras are used for
RGB-NIR cameras are used wherever a scene must be understood in both light and dark, or where NIR reveals something color cannot. The main fields are security, vehicles and biometrics, plus agriculture and industrial inspection.
Security, vehicles and biometrics
Surveillance is the classic case. One camera records in color by day, then relies on invisible IR illumination at night, with no filter that can jam. Automotive interior monitoring is the fastest-growing case. Driver and passenger monitoring needs a clear view of eyes, posture and seat belts in both bright sun and a dark cabin, and regulations are pushing vehicles toward it. Sony's in-cabin sensor uses a global shutter in NIR mode to catch fast movements such as blinking.
Biometric recognition and access control also benefit. NIR images depend less on ambient lighting, so faces look more consistent between a dark lobby and a bright entrance.
If your scene goes from bright to dark in a day, you need NIR in the same frame as color.
Agriculture, inspection and sorting
Healthy plants reflect NIR strongly and absorb red light. The standard vegetation index uses that gap: NDVI = (NIR − Red) / (NIR + Red). A leaf reflecting 50% of NIR and 8% of red gives (0.50 − 0.08) / (0.50 + 0.08) = 0.42 / 0.58, or about 0.72. Stressed plants score lower.
Reliable NDVI needs calibrated bands, so a multispectral design usually suits this job better than a basic RGB-IR module. In factories, NIR also helps show moisture and bruising differences that look identical in color, which supports sorting and quality checks.
How to choose an RGB-NIR camera for your project
Choose by working backward from your lighting: decide on the illumination wavelength first, then check sensor sensitivity at that wavelength, color tuning, lens, shutter and interface. A camera that looks right on a spec sheet can still fail in your lighting if one of these is wrong.
Specs to check before you buy
Work through this list with any supplier, and ask for sample images, not only numbers.
Illumination wavelength. 850 nm gives the sensor more signal but shows a faint red glow. 940 nm is invisible, but silicon responds less there. Ask for quantum efficiency at your wavelength. Sony quotes 35% at 940 nm for its 2.1 µm pixel IMX775, a useful reference point.
Color accuracy in sunlight. Request samples under sun and incandescent light, where NIR leakage shows up.
Lens and filters. Look for IR-corrected lenses, because focus shifts between visible and NIR. A dual-band filter can pass only the bands you want.
Shutter type. Choose a global shutter for fast subjects such as driver monitoring. A rolling shutter is fine for slow scenes.
Interface and drivers.UVC over USB works without custom drivers. MIPI CSI-2 suits embedded boards that need low latency.
Getting color and NIR from Vadzo hardware
Vadzo offers several routes. The Innova-821CRS (from $400) is a 4K GigE camera on the onsemi AR0821 sensor with an auto-switch IR-cut filter, which gives you the day/night design from the table above. For the two-camera route, the Falcon-2020MRS is a 20MP monochrome USB camera with enhanced NIR response at 850 nm and 900 nm, which you can pair with a color module.
Vadzo's engineers also build custom NIR and color LED array boards, add electro-mechanical lens filter control, and select sensors from onsemi, Sony and Omnivision. If you need a single-sensor RGB-IR design, describe your lighting to the application engineering team and they can scope it. There is no minimum order quantity, so you can evaluate before you commit to volume.
Cost and legal questions
Cost depends on the design. Embedded RGB-IR modules sit at the low end. Two-camera rigs cost more, and snapshot multispectral instruments cost the most. Depth cameras are priced as a separate category, so do not compare them directly.
On legality, an NIR camera is an ordinary silicon camera, and owning one is legal in the US. The rules that matter govern what you record and where, such as privacy, consent and workplace laws. Thermal imaging is a different technology with its own rules. This is not legal advice, so check your state's rules before deploying in places where people expect privacy.
Picking the right camera for your project
An RGB-NIR camera gives you color in good light and infrared in poor light from one device. The real choice is between a single RGB-IR sensor, a switchable filter, two cameras or a multispectral design. Match that choice to your lighting, resolution needs and the color accuracy you must keep in sunlight.
Check the NIR sensitivity at your wavelength, ask for sunlight samples and confirm the lens is corrected for infrared. If you want help scoping the hardware, see Vadzo's camera modules and custom engineering options and describe your lighting to the team.
