Open app

Motion recording triggering on nothing

very common

“Customer's getting four hundred alerts a night and it's all headlights and bugs. Now they've stopped looking at any of them, which means the system might as well be off.”

Likely causes

  • Pixel-based motion detection used where object classification is available

    most common

    Events cluster on wind, rain, headlight sweep and moving shadow. If the camera or VMS supports person and vehicle classification and it is switched off, that is the answer. The alert list is its own evidence, because it contains everything and therefore nothing.

  • Insects and webs in the near field, lit head-on by on-axis IR

    common

    Events run almost entirely dusk to dawn, each clip a big bright out-of-focus blob within a metre of the lens; daylight is quiet. The array is not attracting them: most flying insects see UV and short-wavelength visible light and are largely blind to 850 nm. They gather at the housing for its warmth and at whatever visible lighting is nearby, and spiders build where insects already are. The array is the problem because it is coaxial with the lens, so anything in the near field comes back saturated.

  • Detection zone left at full frame, covering sky, road and foliage

    less common

    Open the zone overlay. If it is the default full-frame mask, stop looking. A public road in frame buys you an event for every passing car, and the customer never asked to monitor the road.

  • Sensitivity turned up during commissioning to prove it worked

    The event log contains triggers where the clip shows no visible change at all. Somebody wound it up to demonstrate the system and never wound it back.

  • Auto-exposure and day/night switching triggering the detector itself

    Events land precisely at dawn and dusk and at the moment the image flips to mono. The clip shows the whole frame changing brightness rather than anything moving.

  • Moving shadows, wet tarmac reflections, water, or a flag in frame

    Event frequency tracks weather and time of day rather than activity. Watch ten clips back to back: a repeating physical object gives itself away immediately.

  • Camera physically moving: loose mount, pole sway, HVAC vibration

    Whole-frame motion events with a slight positional shift between clips, correlating with wind or plant cycling. Grab the mount and check for play.

  • Not false alarms at all: no pre-buffer, so real events look useless

    Different fault, identical complaint. Clips start mid-action and the customer calls them worthless. Check pre-event buffer and post-event duration before assuming the detector is wrong.

What to bring

  • VMS or camera admin with event log export
  • Detection zone overlay access
  • Ladder or lift
  • Lens cloth
  • External IR illuminator and mount if relocation is the fix

Safety

Checking the mount for play and relocating an illuminator are both at-height tasks. Set the platform under the work rather than reaching from it. Relocating an illuminator is a new mount in its own right: it needs its own fixings sized for the structure, and its own bonding consideration on metal. Do not look into an active IR array at close range: near-IR produces no blink or pupil response, so nothing warns you that you are overexposing your retina. Cut power to the illuminator before working in front of it.

Steps

  1. Watch twenty consecutive event clips, oldest first

    Do not touch a setting until you have seen the pattern. Twenty clips tells you whether this is insects, headlights, weather, exposure switching or a moving camera. You cannot deduce any of that from the event count.

    If that doesn’t do it

    Clips are inconclusive: check whether the events even have video attached. If they do not, this is a recording or trigger-routing fault, not a tuning job.

  2. Look at the time-of-day distribution of the events

    All night means the near field and the IR. Spikes at dawn and dusk mean exposure and day/night switching. Business hours means real activity or a public road in frame. Windy days only means foliage or a loose mount.

    If that doesn’t do it

    Evenly distributed across the clock: go to zone and sensitivity.

  3. Open the zone overlay and cut it back to what the customer actually cares about

    Exclude sky, public road, neighbouring property, foliage and any water. Exclude the near field where insects fly. The zone should cover the ground plane a person would walk across, not the whole picture. Free, immediate, and the single biggest win available.

    If that doesn’t do it

    Zone is already tight: enable classification.

  4. Turn on object classification if this model and VMS support it

    Person and vehicle classification removes the overwhelming majority of weather and insect triggers because a blob now has to look like an object to count. Capability, feature naming and licensing all vary by brand and by model generation. Check what this exact model and VMS version supports rather than assuming parity across the fleet.

    If that doesn’t do it

    No classification available on this model: tune object size and dwell.

  5. Set a minimum object size and a dwell time

    A target has to be big enough and persist long enough to count. Object size is expressed as a fraction of frame on most platforms, which means the same numeric value means completely different things on a 4 mm and a 12 mm lens. Set it per camera against its own field of view and range.

    If that doesn’t do it

    Still noisy at night: deal with the illuminator.

  6. Move the IR illuminator off the camera axis

    This is the only fix that removes the cause instead of filtering the symptom, and it usually improves the night image at the same time. Classification and near-field zone exclusion suppress the alerts and are worth doing regardless, but the blobs are still in the picture. Mount the illuminator well clear of the camera and aim it across the scene rather than along the lens axis.

    If that doesn’t do it

    Illuminator cannot be relocated: schedule bubble cleaning, raise the threshold, and tell the customer why.

  7. Check the mount and the day/night switching hysteresis

    No play in the mount, no sway, no arm resonating with a rooftop unit. Then widen the day/night hysteresis or set fixed schedules so the camera is not flipping modes repeatedly at dusk.

    If that doesn’t do it

    Mechanically solid and hysteresis set: baseline and hand over.

  8. Baseline for 48 hours and report the new event count

    Write down the before and after. Four hundred a night down to nine is what stops the callback and settles the argument about whether the system works. Without the number, you have nothing to show for the afternoon.

References

  • IEC 62676-4
  • IEC 62471: photobiological safety of lamps and lamp systems

More Video Surveillance guides

All 54 guides