ToF camera outdoor sunlight: why the point cloud goes fuzzy

A time-of-flight camera that scans cleanly indoors can fall apart the moment it faces open sky: surfaces lose definition, edges smear, and the point cloud reads as fog. The camera has not failed. The sun has joined the measurement as a second, unmodulated infrared source, orders of magnitude brighter than the light the camera emits. This guide is for anyone putting a ToF 3D camera at an outdoor or dock-side station — pallet detection, trailer profiling, yard logistics. It gives you the saturation mechanism with real numbers, the tests that confirm sunlight is the cause, the settings that recover usable depth, and the point at which shading or a different sensor technology is the only answer.

What outdoor failure looks like

The classic progression: the same pallet scans sharp inside the warehouse, then comes back fuzzy and poorly defined on the yard, with noise across the whole frame rather than at a few edges. Three details separate sunlight from focus, motion or range problems:

• The degradation tracks the sky. A re-capture at dawn, at dusk or under a tarp sharpens the same view with no settings touched.
• Sunlit faces suffer first. Shaded sides of the same object hold together far better.
• The amplitude or confidence image sags wherever the sun lands, before the depth view looks obviously wrong. Learn to read that image early — decoding confidence image bit flags shows how.

If the fuzz stays the same day and night, and moves with the part instead of the sky, look at motion, range or surface instead; the cause table below still helps you rule suspects out in order.

Why sunlight saturates a ToF receiver

A ToF camera emits its own modulated light in the near infrared — 850-940 nm is the working band for most designs — and computes depth from the round-trip time or phase of the return; the Wikipedia time-of-flight camera overview covers the principle. The receiver is built to measure light that carries the camera's own modulation and to ignore everything else.

Sunlight defeats the ignore step. About half of the sun's ground-level power arrives in the infrared, and in the 850-940 nm band the clear sky delivers roughly 0.7-1.0 W/m²/nm of spectral irradiance. The optics pass a band tens of nanometres wide around the emitter wavelength, so a steady, continuous in-band flood reaches the sensor in every frame.

That flood charges every receiving pixel toward its well limit. The modulated swing you actually measure is a small difference on top of that pedestal; once the wells fill, the swing has nowhere to live, depth precision collapses, and the cloud fuzzes out. Signal-to-noise ratio is the entire story: the sun attacks it in-band, through the same filter your own light passes. The mechanism is identical to the one behind depth camera noise in bright ambient light — sunlight is simply that problem at full strength.

The power that matters follows from three quantities:

Worked example: in-band sun power through a 25 mm lens at one sun

Inputs. Clear-sky irradiance at ground level: 1000 W/m², the standard one-sun level used across solar engineering. Infrared share: about half, so roughly 500 W/m² lands as infrared. Spectral irradiance at 850 nm: 0.85 W/m²/nm, a mid-range value from the AM1.5 reference spectrum (the range runs 0.7-1.0). Bandpass width: 60 nm around 850 nm, a typical ToF filter width and the one input worth checking for your model. Aperture: a 25 mm front element, a common size on these cameras, so A = π × (12.5 mm)² = 491 mm² ≈ 4.91×10⁻⁴ m².

Rule in words. Multiply the spectral irradiance by the filter width to get the in-band irradiance; multiply that by the aperture area to get the continuous power delivered through the optics.

Arithmetic.

• In-band irradiance: 0.85 W/m²/nm × 60 nm = 51 W/m²
• Power through the lens: 51 W/m² × 4.91×10⁻⁴ m² = 0.025 W ≈ 25 mW

Result. Direct sun pushes about 25 mW of continuous in-band infrared through the optics, every frame. Without a bandpass filter the same aperture collects 500 W/m² × 4.91×10⁻⁴ m² ≈ 250 mW. Heavy overcast cuts the sky to roughly a tenth of one sun, which still leaves about 2.5 mW. For scale, the return you are trying to measure is tiny: an emitter of order one watt of optical power, spread into a wide cone, bounced off a diffuse target a few metres away and captured again through the same aperture, arrives in the microwatt range. Direct sun is tens of thousands of times stronger — exactly why the wells fill.

Decision. If direct sun reaches the measured surfaces, tuning is a stopgap: shade the cell, schedule the work out of the bright hours, or change technology. If the sun only enters the field of view or glints off a few surfaces, a hood or a re-aim clears that path outright. If the cell sits in shade or the task runs at dusk or night, the ToF camera remains the right tool.

Rank the causes and test each

Outdoor fuzziness has four usual suspects. Work down the table in order; each test costs minutes.

Cause 3 hides behind cause 1: a camera standing in direct sun often shows both, and the heat problem persists after a cloud passes. The thermal derating guide for 3D cameras walks through separating them.

Fixes in daylight, ranked by payoff

1. Shade the measured volume, not just the camera. The dominant term in the worked example is sunlit surfaces scattering in-band light into your lens. A canopy, a dock recess, a shrouded window, or a mount position that keeps the measured volume in shadow cuts the continuous 25 mW flood toward the 2.5 mW overcast level at a stroke. It is the only fix that removes the cause at one sun.

2. Keep the sun out of the field of view. Re-aim so the disc and its specular reflections never enter the frame, and fit a lens hood to block off-axis rays. A hood costs nothing and removes the direct path completely; it does nothing against scatter off the target, so pair it with fix 1.

3. Cap the exposure. Shorten the integration time until the amplitude histogram stops crowding its upper limit. The trade: collected light falls in proportion, and depth noise grows roughly as one over its square root, so halving exposure raises noise about 40 percent. The ToF exposure time and frame rate trade-off covers the other side of that trade. Shorter range and brighter surfaces move the same lever in your favour.

4. Filter on amplitude or confidence. Reject pixels whose amplitude or confidence falls below a threshold set during a noon test. You lose coverage of the worst patches, but the surviving cloud is trustworthy — and the threshold doubles as a production gauge: a station that passes at dawn and starts failing mid-morning is telling you the sun is winning.

5. Buy for the band. None of the above changes wavelength. If the station must live outdoors, select at purchase: 940 nm-class emitters sit in a water-vapour absorption dip of the solar spectrum and start with less in-band sun than 850 nm designs. The guide to iToF modulation frequency and ambient light explains how the modulation and wavelength choices interact with ambient light.

The sensor decision for each outdoor case

Match the technology to the station before spending a day on settings:

Indoor pallet detection. No sun, no saturation problem; pick on range, field of view and software. Any ToF or structured-light camera in the 3D sensors category fits.

Shaded dock, covered yard, dusk-to-night schedule. ToF stays the right tool once tuned. For controlled-light full-frame depth at logistics range, a ToF sensing camera such as the MV-DT01WMU does the job.

Pallet and parcel volume under a roof. A volume-class structured-light camera fits the task: the structured-light camera for pallet and parcel volume measures pallet and parcel volume at logistics range in covered yards.

Open-sky direct sun, full-frame 3D. No flood-emitter technology wins at one sun. Build shade over the measured volume and schedule around the bright hours, or change what you measure. For profiles — edges, gaps, flatness, channel and truck-bed geometry — laser profilers are the standard outdoor instrument: the measurement is a fast laser-line reflection read in microseconds, not a continuous integration against the sky. The DP4000-series MV-DP4180-03P is one such profiler.

The middle ground. Other families worth comparing for shaded stations and profile work:

Field checklist before you commit the station

• Survey at the brightest hour. Stand at the station at solar noon on a clear day and map where direct sun falls across the seasons; a site that scans in March can fail in June.
• Keep the sun's daily path out of the field of view. Re-aim, fit a hood, then check again at low morning and evening sun angles.
• Shade the measured volume if the process allows. That is what moves a station from the 25 mW case toward the 2.5 mW case.
• Record the amplitude or confidence image in every trial. Clipping appears there before the depth view degrades.
• Set exposure and amplitude thresholds at the brightest hour. Weather on commissioning day is not a specification.
• Mount for wind and heat. Rigid fixturing stops smear; shading or airflow stops thermal derating.
• Close the housing question separately. Rain, dust and washdown are their own engineering problem — the outdoor camera enclosure IP rating guide covers what a bare camera needs outside. If operators will read a screen in the same cell, choose a sunlight-readable HMI for outdoor industrial applications rather than a standard panel.

The short version: shade or schedule around direct sun, cap exposure and set amplitude thresholds for the brightest hour, and move to a laser profiler when open-sun profile measurement is the task. To match hardware to a station, filter the camera finder, work through the system configurator, or compare technologies in the industrial 3D camera selection guide; the 3D topic page collects the rest of the guides.