UV reflective imaging with a mono camera: light, filter and lens

Sunscreen turns black, glass goes opaque, invisible inks and residue maps snap into view: those are UV-A reflective effects, and a standard mono machine-vision camera can capture them. The catch is that silicon's response at 365 nm is a small fraction of its visible response, so every stray visible photon competes against a weak signal. This guide gives you the three decisions that make UV reflective imaging work: which lamp wavelength to buy, which filter isolates the band, and which optics transmit it, ending with a worked example that sizes field of view, diffraction and depth of field for a hand-sized subject.

What a mono sensor sees at the UV-A band

Silicon converts UV-A photons into charge, but weakly. Published measurements of back-side illuminated CMOS pixels show quantum efficiency falling by nearly 50% between 400 nm and 200 nm, and reaching 35-50% across 240-400 nm only in UV-enhanced scientific detectors. An ordinary front-illuminated machine-vision sensor sits below those figures in this band: usable for contrast, but a fraction of its own visible sensitivity.

Two more losses sit in front of the photodiode. The sensor window absorbs a slice of the UV, and on a color camera the Bayer dye array absorbs most of it, so mono is the right sensor here. A board-level mono module with no IR-cut filter passes UV readily, and passes visible and near-IR just as readily, which is the problem the next section solves. Before you remove or swap a window, see what a protective filter window changes for a vision lens.

The camera decision: mono sensor, a window that transmits at the lamp wavelength, and an exposure budget far beyond the microseconds of visible strobing.

The filter decides whether the effect appears

The effect lives only in the UV band. Silicon's response climbs from the UV toward visible wavelengths, so a few percent of violet leaking through a filter can carry more charge than all of the reflected UV light returning from the part. Room lighting, the lamp's own violet tail and IR leak each pile on top. The camera-side filter is not an accessory: it defines the image.

The options, strongest first:

1. Dedicated UV-pass filter. An interference filter built for UV-A reflective photography, often listed as a UV Black type. It passes the UV-A band and blocks above it. This is the choice whenever the room is not dark.
2. Astronomy-grade UV filter. The deepest visible rejection available; more than a production demo needs, but the reference standard.
3. Violet glass over NIR-blocking glass. The budget stack: a violet number-47 filter, which passes violet and UV but leaks near-IR, over BG39 glass, which passes UV and blue and blocks near-IR. Stacked, they pass UV-A plus some violet and remove the IR leak. Workable inside a dark enclosure where your lamp is the only source.
4. Single violet glass. Sold as UV pass, but it leaks enough violet to wash the effect out. Not recommended alone.
5. No filter. Visible and IR dominate and the effect disappears.

If you stack two budget glasses, their transmissions multiply:

Each extra glass squares the leak but also cuts throughput, lengthening exposure. A dedicated UV-pass filter wins both ways.

Decision: dark enclosure with only your lamp, the violet-plus-BG39 stack is enough. Open room, mixed lighting or repeatable production, a dedicated UV-pass filter, and the deeper its stated visible blocking, the cleaner the image.

Reading a filter sheet

Three lines matter. The transmission curve must cover the lamp wavelength with margin, so buy against the curve, not the name. The visible blocking should be stated as a rejection level above 400 nm; single-glass filters rarely state one, which is itself the tell. And the clear aperture must cover the lens aperture at your working f-number.

The lamp: wavelength first, power second

Buy by peak wavelength. The classic skin, print and residue effects sit around 365 nm, the Wood's lamp band. Many low-cost UV torches are actually 395-405 nm violet LEDs: silicon sees violet well, so images come easily, but effects tuned to that band weaken and violet slips through budget filters far more readily. Read the wavelength on the lamp's sheet, not the glow. Where UV sits among the other vision bands is mapped in wavelength ranges of light in machine vision.

Irradiance falls with the square of distance. Moving a lamp from 100 mm to 300 mm cuts light to one ninth and multiplies exposure by nine. Mount the lamp 100-200 mm from the subject for a demo rig; if mechanics force a longer stand-off, specify a higher-output bar or overdrive it, because a short strobed pulse at high current recovers signal without blur, as covered in strobe overdrive, duty cycle and timing.

The special lights category covers UV-colour lights among AOI, strobe and structured types, plus a customized series built to your geometry.

Lenses that transmit UV

A camera-side filter cannot add transmission the lens lacks. Standard machine-vision lenses are specified for visible light, their transmission at 365 nm is uncontrolled, and every point lost there lands on top of the weak sensor response. Dedicated UV lenses state the band directly: the LS-SWL-0291 is a 25 mm, F2.0 C-mount lens covering 240-1000 nm on a 2/3" format with an 8-11 mm image circle and 6-10 MP resolution.

Longer stand-offs and larger sensors need longer focal lengths or bigger image circles. The LS-SWL-0269 is a 60 mm UV lens for APS-H sensors with a 30-45 mm image circle in EF or C-mount, and the LS-SWL-0280 is a 1.4x UV converter covering 190-700 nm for a 4/3" sensor with a 16-22 mm image circle. All three live in the thermal, SWIR and UV lenses category.

Values as printed in each product's data sheet; open a part for its drawing and full specifications.

Whatever the lens, screw the UV-pass filter onto its front thread and keep any extra window out of the path unless it is specified to transmit UV. What changes outside the visible band, from coatings to focus shift, is explained in SWIR lens and UV lens basics, and the habit of testing contrast before you commit is laid out in reflected UV surface inspection.

Worked example: framing a hand at 300 mm with a 25 mm UV lens

Inputs:

• Sensor: 2/3" mono, 5 MP (2448 × 2048 pixels), active width 8.8 mm; the standard 2/3" format is 8.8 × 6.6 mm.
• Lens: LS-SWL-0291, 25 mm focal length, F2.0, 240-1000 nm band.
• Working distance: 300 mm, chosen to clear the lamp housing and frame a splayed hand with margin.
• Lamp: 365 nm peak, matching the effects band.

The rules, in words: field of view equals sensor width times working distance divided by focal length; magnification is sensor width divided by field of view; the diffraction spot is 2.44 × wavelength × f-number; depth of field is approximately 2 × f-number × pixel size × (1 + m) / m².

Arithmetic:

• Field of view: 8.8 mm × 300 mm / 25 mm = 105.6 mm wide (about 106 mm) and 6.6 mm × 300 / 25 = 79.2 mm tall (about 79 mm).
• Pixel pitch: 8.8 mm / 2448 = 3.6 µm, so sampling is 105.6 mm / 2448 = 43 µm per pixel.
• Magnification: m = 8.8 / 105.6 = 0.083.
• Diffraction: 2.44 × 0.365 µm × 2.0 = 1.8 µm at F2.0, half a pixel; at f/4 it grows to 3.6 µm, one pixel, still acceptable.
• Depth of field: 2 × 2.0 × 0.0036 mm × 1.083 / 0.0069 = 2.2 mm at F2.0; at f/4 it roughly doubles to 4.5 mm.
• Light budget: at a 300 mm stand-off a lamp rated at 100 mm delivers one ninth of its irradiance, so exposure rises about 9× unless the lamp moves close or overdrives.

Result: a 25 mm F2.0 UV lens at 300 mm frames a 106 × 79 mm field with 43 µm per pixel, a hand with margin, and millimeter-scale marks resolve trivially. Diffraction is never the limit between f/2 and f/4 here. Depth of field is 2.2 mm at F2.0, so fix the part plane at a set distance, or stop to f/4, pay about 4× exposure and hold 4.5 mm of depth. If you need better than 43 µm sampling, halve the working distance to 150 mm: the field drops to 53 × 40 mm and sampling to 22 µm per pixel.

Bench test before you buy

• Smear sunscreen on a matte black card. Through the filter the smear must sit near-black while the card stays dark; if the filtered scene still looks like ordinary light, the filter leaks visible.
• Image a glass pane. With a working filter it should fall to dark grey, because only the weak UV surface reflection remains.
• Lock exposure manually once tuned. Auto modes hunt on a mostly black scene and change the image between frames.
• For long exposures, take a dark frame, lens capped at the same exposure, and subtract it to expose the real pattern noise.
• Confirm the lamp's peak wavelength on its sheet before blaming the camera.

When the usual answer is wrong

The subject fluoresces instead of reflecting: security threads, some adhesives and many minerals glow visibly under the UV-A lamp. Then the camera filter flips, because you block the excitation with a long-pass filter and image the emitted visible light in a darkened cell. The choice between the two modes is set out in reflected UV vs UV fluorescence imaging, and the excitation side in fluorescence inspection lighting.

You need measurements, not contrast: quantitative UV radiometry is not what a machine-vision mono sensor is for; the UV-enhanced scientific detectors quoted earlier exist for that job. And a 254 nm short-wave lamp is a different project entirely: ordinary glass optics and windows absorb that band completely, so it takes quartz optics and the strictest safety regime, not a filter swap.

Checklist before ordering

Lamp:

• Peak wavelength 365 nm stated on the sheet, not violet 395-405, because both the effects and the filter need that band.
• Mounting plan at 100-200 mm from the subject, or an overdrive-capable driver for longer stand-offs.

Filter:

• Transmission curve covering the lamp band, stated visible blocking, clear aperture covering the lens aperture.
• Violet-plus-BG39 stack only where the set-up is a dark enclosure.

Camera:

• Mono sensor, no Bayer array, window transmitting at the lamp wavelength.
• Manual exposure control and dark-frame subtraction in the software plan.

Optics:

• A lens specified for the band, focal length sized by the field-of-view arithmetic above.

Next step: size the optics in the lens finder or the lens configurator, pair them with a UV light from the lighting finder, and for a built-to-print lamp or a full camera-lens-light match use the contact page. More lighting write-ups sit on the lighting topic page.