Ailytics raises US$7.4M Series ARead more

Comparing Forklift Pedestrian Detection Systems: Onboard Cameras vs. Existing CCTV

Forklift pedestrian detection: onboard cameras cover the truck's blind spots; site CCTV analytics cover the full floor. Coverage map, verdict, and procurement checklist for US warehouses.

Last updated

18 min read

Summary

  • Onboard vehicle cameras protect the truck by alerting the driver to pedestrians in blind spots, while site-level CCTV analytics protect the floor by detecting risks across existing camera coverage.
  • Coverage, not accuracy, is the deciding factor: site-level analytics covers everyone the cameras can see, with no wearable, while tag-based systems miss untagged contractors, visitors, and carrier drivers.
  • Choose onboard cameras for tight aisles, reversing docks, and confined areas with no useful fixed-camera angle; choose site-level analytics for mixed-population floors, dock doors, and aisle intersections.
  • Neither technology replaces physical separation, and both camera approaches struggle behind racking. In 2024, 84 workers died in incidents involving forklifts, order pickers, or platform trucks, and pedestrian-involved nonfatal cases had a 20-day median recovery.
  • Ailytics adds site-level pedestrian detection to existing IP cameras via RTSP/ONVIF, with up to 99% accuracy and deployment in under two weeks.

The near-miss at the dock during shift change is the one nobody writes down. A carrier's driver cuts through the receiving lane, a forklift reverses out of a bay, and the two come within a few feet of each other. Nobody was hurt. Nothing gets logged. The site's near-miss report shows zero for the month.

Forklift pedestrian detection exists to close that gap. The harder question is which system closes which gap. Onboard cameras and site-level CCTV analytics are not competing answers to the same question. They protect different people.

Quick Verdict: Which System Is Right for Your Site?

Onboard cameras protect the truck. Site-level CCTV analytics protect the floor. That distinction decides the purchase.

Onboard vehicle cameras are mounted on the forklift, detect people in the truck's immediate blind spots, and alert the driver directly on PPE violations. They are bought per-truck and scale with your fleet. Site-level analytics run on your existing fixed cameras, cover every person in camera view regardless of what they are wearing, and alert a supervisor or safety manager. They are bought per-site.

Most buyers treat accuracy as the deciding criterion. It is not. Both approaches can achieve high detection accuracy. The real question is who each system can see. A contractor who walked in from the dock, a carrier's driver cutting through a receiving lane, an auditor on a facility tour: these people are invisible to tag-based proximity systems unless they were issued a badge. They are visible to any camera-based system, onboard or fixed, as long as there is a clear line of sight.

Busy US sites running mixed populations, active dock doors, and high-traffic intersections generally need to close two different gaps. This guide helps you decide which gap to close first.

What an Onboard Camera System Actually Detects

An onboard forklift pedestrian detection system consists of one or more cameras mounted directly on the vehicle. The AI processes the feed in real time and classifies human shapes within the camera's field of view.

The alert goes to the driver, not to a control room. The design intent is to put the warning in front of the person who can act on it immediately. Some systems can be integrated with vehicle controls to trigger a slowdown or stop, though that depends on the forklift model and controller.

Coverage is defined by the camera's angle and the obstructions in front of it. The system detects people the camera can see. It does not detect people behind racking, around a blind corner, or anywhere outside the field of view. Detection approaches differ by what they detect and what the pedestrian must carry. Onboard cameras require nothing from the pedestrian, but they require a clear sightline.

The procurement unit for onboard systems is the truck. Every forklift in your fleet is a separate line item.

What Existing CCTV Analytics Detects

Site-level analytics adds an AI layer to the IP cameras already mounted on your ceiling, walls, and dock areas. The software connects over RTSP and ONVIF, the standard protocols cameras and recorders from different manufacturers use to communicate. No new hardware goes on the forklifts.

The system tracks people and vehicles across camera coverage and fires an alert when a defined condition is met, such as a person entering a vehicle-only zone while a forklift is present and moving. The alert goes to a supervisor or safety manager, typically with a short video clip. That person decides the response.

Because detection is visual and requires no wearable, the system covers everyone the cameras can see: permanent staff, agency workers, contractors, delivery drivers, visitors on a facility tour. That matters on sites with high personnel turnover or frequent third-party access.

The procurement unit is the site. Cost scales with the number of camera feeds analyzed, not with fleet size.

The Difference Is Not Accuracy. It Is Who Gets Detected.

Both onboard cameras and site CCTV analytics can perform at high accuracy. The system that cannot see a person at all is not running at low accuracy. It is simply not covering that person.

Take the people a US warehouse or distribution center floor actually holds. Permanent forklift operators and permanent warehouse staff on foot are covered by all three approaches. An onboard camera sees them when they are in that truck's view, a site camera sees them anywhere in camera coverage, and a tag sees them if they were issued one and put it on.

The picture splits on the next four groups. Agency and temp staff, contractors doing maintenance, cleaning or IT work, carrier and delivery drivers crossing the dock, and visitors, tours and auditors are all covered by both camera approaches on the same condition: they are in the camera's line of sight. A tag-based system covers none of them unless someone issued a badge and the person is wearing it. That is the whole difference for these groups, and it is the one that matters most on a floor with high turnover or constant third-party access.

The transient population problem. Agency and temp workers, contracted cleaning and maintenance crews, carrier drivers at the dock, and anyone on a site visit typically account for a significant share of foot traffic at peak periods. A tag system's protection for all of them is only as complete as the tag-issue process, which is rarely airtight. Camera-based systems, both onboard and fixed, cover these people by sight with no process dependency.

The blind corner problem. Every camera-based system, whether mounted on the truck or on the ceiling, is limited by line of sight. A person behind racking or around a blind corner is invisible to an onboard camera, and a site camera sees them only if a camera happens to hold an angle on that corner. This is where tag-based systems using ultra-wideband (UWB) ranging hold a genuine advantage: UWB signals pass through physical obstructions, so the system detects tagged personnel even when the camera cannot. If most of your risk sits in dense racking aisles where workers are frequently behind shelving, a camera-only strategy leaves a gap that UWB closes.

Tag Systems Miss Most of Your Floor

Where Onboard Cameras Are the Right Call

Onboard vehicle cameras make the most sense when risk is specific to one truck's immediate environment and the most useful intervention goes directly to the driver.

Dense racking and tight aisles. In narrow-aisle configurations, the operator's field of view changes constantly. Sightlines close and open as the truck moves. A rear-mounted or side-mounted onboard camera gives the driver real-time awareness of what the mirrors do not show.

Reversing in confined docks. Reversing is one of the highest-risk maneuvers on a busy floor. A camera with pedestrian detection mounted at the rear of the truck addresses the physics of the movement: the operator cannot see behind the load, and a rear-facing system does.

Confined spaces with no viable camera angle. Some areas of a facility cannot be covered effectively by a fixed ceiling camera: low-clearance zones, outdoor yards with variable lighting, temporary work areas, or sections with structural obstructions. Where a fixed camera cannot be positioned to see the floor clearly, an onboard system travels with the risk.

Driver-direct intervention. When the right response is an immediate in-cab warning, onboard is the correct architecture. The alert reaches the person controlling the vehicle, with no delay for a supervisor to receive a notification and respond.

Federal operator training rules require pedestrian traffic to be covered for the areas where the truck actually operates (29 CFR 1910.178(l)(3)(ii), under workplace-related training topics). Training is required. It is not a control. An operator aware of a pedestrian policy is not the same as an operator who receives a real-time alert when a person enters the zone behind the truck. Onboard detection closes that gap directly.

Where Site-Level CCTV Analytics Is the Right Call

Site-level analytics on existing CCTV is the right choice when risk isn't contained within one truck's operating area but is distributed across the floor: intersections, dock doors, and the crossings between zones.

Mixed-population floors. Where contractors, agency workers, and carrier drivers move through the facility regularly, a detection layer that requires no wearable is the only one that covers them reliably. There is no tag to issue, no compliance to enforce. Anyone the camera can see is detected.

Dock doors at shift change. Congestion at dock doors during shift change is among the highest-density forklift-pedestrian interaction points on any busy site. Multiple vehicles, multiple people, tight timing. A fixed camera covering that entrance can log every close interaction, not just the ones involving a specific truck.

Aisle intersections. Sites report that compliance with "stop and sound the horn" rules at obstructed intersections is difficult to verify without observation. Operators are required to slow down and sound the horn at cross aisles and other locations where vision is obstructed, and to travel with the load trailing where the load blocks the forward view (29 CFR 1910.178(n)(4)). Enforcement depends on seeing the behavior. A fixed camera at an intersection logs compliance and near-miss events automatically, and that data feeds directly into coaching and toolbox talks.

Turning reactive cameras proactive. On most sites, CCTV footage is reviewed after an incident, not before one. The cameras recorded the event, but nobody was watching. Site-level analytics turns existing hardware from forensic to preventive, with no hardware installation on a single forklift.

What Neither Technology Solves

Detection is not separation. The OSHA guidance on pedestrian traffic is explicit: forklift traffic should be separated from pedestrians wherever possible. Designated walkways, physical barriers, guardrails, and controlled entry points at intersections are engineering controls. A detection system is a warning layer on top of those controls, not a substitute for them.

OSHA publishes no numeric safe-distance rule and no federal speed limit for pedestrians working around forklifts. The "stay 10 feet back" or "three truck lengths" figures that appear in training materials and signage are employer and insurer conventions, not federal requirements. The obligation under the General Duty Clause is to provide a workplace free from recognized hazards. How that obligation is met is the employer's determination.

The person behind racking. Both onboard and site-level camera systems require a clear line of sight. A worker who steps behind a rack while a forklift rounds the corner is not visible to either. Physical separation handles this case. Where layout constraints make separation impossible, a UWB tag system addresses it because the signal passes through the obstruction. That is the limit of any camera-only comparison.

Near-misses that never get logged. Detection systems make near-misses visible, but only if someone responds to the alert and the event is recorded. Sites consistently find that their near-miss log understates actual close-call frequency before they start systematic monitoring. Federal data on forklift incidents supports the severity argument: 84 workers died in 2024 in incidents involving forklifts, order pickers, or platform trucks. Pedestrian-involved nonfatal cases carry the longest median recovery at 20 days away from work, against 8 days for all injury types. That figure represents the cases that were reported. Before systematic monitoring, close calls usually happen at a much higher rate than the written log shows.

Alert fatigue. Honking becomes white noise on a busy site. A detection alert that fires too often for low-confidence reasons reaches the same fate. A system configured with compound rules, such as confirmed movement plus confirmed pedestrian presence plus time in zone, keeps alerts meaningful. A system that fires on every bounding-box overlap in a 2D image does not.

See It Detect. On Your Cameras.

The Procurement Questions That Actually Decide It

These questions separate a useful evaluation from a vendor demo cycle.

What unit are you buying? Onboard detection is a per-truck purchase. Every forklift added to the fleet is a fresh line item. Site-level analytics is a per-site purchase, scaled to the number of camera feeds analyzed. That structural difference shapes the total investment more than any unit price.

Who owns the footage and how is it stored? For site-level systems, where is data processed: on-site, at the edge, or in the cloud? Can faces and bodies be blurred while still preserving the safety event data? In unionized environments and sites with active privacy policies, this question surfaces early.

How does the system handle shift change and agency staff? An onboard system covers whoever is in its field of view regardless of headcount. A site-level system does the same within camera coverage. The question is whether camera coverage actually extends to the areas where the highest-turnover personnel work.

What is the false-alert behavior? An onboard system that triggers on a pallet or doorpost trains the driver to ignore it within a week. A site-level system that fires on perspective overlap in a 2D image generates noise that supervisors stop responding to. Confirm exactly what compound conditions must be met before an alert fires.

What response does the site need? Options range from an audible cab warning, to a supervisor notification with video, to a slowdown or braking intervention on the vehicle. The response type should match the site's risk profile and operational tolerance. Automatic braking integrations have been rolled back on some sites because the slowdown effect on throughput was not acceptable. Confirm what your forklifts' controllers support before specifying a response.

How is near-miss evidence produced? The safety case for any detection system rests on its ability to convert close calls into logged, reviewable data. Confirm that the system produces time-stamped, location-tagged event records that can be used in traffic management reviews, audits, and the refresher training required after an accident or near-miss (29 CFR 1910.178(l)(4)(ii)(B)).

Where Ailytics Fits

Ailytics provides site-level forklift pedestrian detection analytics on your existing IP cameras, connecting via RTSP or ONVIF with no new hardware on the trucks or ceiling.

The detection logic addresses a common failure mode in 2D systems. A flat 2D check fires when a person's bounding box overlaps a vehicle's bounding box in the image frame. That triggers for anyone who merely appears close from the camera's angle, regardless of actual ground-level proximity. Ailytics converts the view into a 3D spatial model, draws the danger zone flat on the ground around the moving forklift, and checks whether the person's feet are inside it. A parked forklift does not raise alerts. The alert fires on a moving machine with confirmed pedestrian presence inside the ground-level zone. Models are fine-tuned specifically for heavy industries and conditioned for harsh operating conditions: variable warehouse lighting, dock glare, night shifts, dust in yards.

Accuracy reaches up to 99 percent. Grace periods, alert cooldowns, and per-object cooldowns keep alerts meaningful. The default response is notification to a supervisor. Where a site specifies a slowdown or brake integration, that is configured as part of the deployment, depending on the machines and site requirements.

Every flagged event is logged with a video frame, timestamp, and location. Those records feed traffic management reviews, audits, and toolbox talks. For logistics and warehousing operations, the outcome is a documented reduction in safety non-compliance of up to 80 percent. Ailytics can go live on existing cameras in under two weeks.

Where workplace surveillance is a live objection, Ailytics can blur faces and bodies while still tracking the interaction, so the safety event survives and there is nothing identifying in the stored clip.

For a broader view of site safety tools, see the best warehouse safety software roundup.

Protect the Truck, the Floor, or Both?

Onboard cameras protect the truck; site-level CCTV analytics protect the floor. They detect different people and answer different questions. Treating them as competing solutions to the same problem produces the wrong purchase.

Choose onboard vehicle cameras when the primary risk sits inside a single truck's blind spots: dense racking, tight aisle navigation, reversing in confined docks, or any area where no fixed camera can hold a useful angle on the floor. This is also the right architecture when the response must reach the driver directly, whether as a warning or as a vehicle intervention.

Choose site-level CCTV analytics when the risk is systemic and distributed across a mixed-population floor: dock doors at shift change, aisle intersections, yards, and any area where contractors, carrier drivers, visitors, and agency workers move through without wearables. This is the correct choice when you need to turn existing camera infrastructure from a forensic tool into a preventive one, and when you need site-wide near-miss data to identify hotspots you cannot yet see.

For the busiest US warehouses and distribution centers, with a large forklift fleet and constant third-party traffic at the dock, the honest answer is often that both gaps are real. Site-level data can show you which trucks and which zones carry the highest interaction frequency, and that analysis informs where an onboard system adds the most value. Neither technology replaces physical separation, and neither closes the behind-racking gap that UWB proximity systems address. The hierarchy of controls still holds.

See how Ailytics turns your existing cameras into a site-wide pedestrian detection layer, and book a demo to map coverage against your floor plan.

Frequently Asked Questions

What is the difference between onboard and site-based forklift pedestrian detection?

Onboard systems are cameras mounted on the forklift that protect its immediate blind spots and alert the driver in real time. Site-based systems use existing fixed cameras to monitor the entire floor, detect interactions between vehicles and pedestrians across the coverage area, and alert a supervisor. They serve different functions and cover different populations.

Do I need new cameras for forklift pedestrian detection?

For onboard systems, yes: a camera assembly is installed on each truck. For site-level analytics, no. Systems that connect via RTSP or ONVIF integrate with the IP cameras you already own. Ailytics, for example, goes live on existing cameras in under two weeks with no hardware installation on the vehicles.

Does pedestrian detection work for contractors and visitors who are not issued tags?

Camera-based systems, both onboard and site-level, detect people visually. They cover anyone in the camera's field of view regardless of what that person is wearing. This is a direct advantage over tag-based proximity systems, which provide no protection to untagged visitors or contractors unless a badge has been issued and is being worn.

Is a camera-based system as accurate as a UWB tag system?

UWB and cameras are strong in different conditions. UWB is more reliable for detecting tagged personnel through physical obstructions such as racking because the signal does not require a sightline. Advanced camera systems such as Ailytics achieve up to 99 percent detection accuracy within camera coverage and use 3D spatial analysis to reduce false alerts. The right comparison is not accuracy against accuracy but coverage against coverage: what each system can see, and under what conditions.

Can forklift pedestrian detection stop the truck automatically?

Some onboard systems offer integration with vehicle controls to trigger a slowdown or stop, subject to forklift model and controller compatibility. Site-level platforms such as Ailytics default to supervisor notification and can be configured for heavier interventions on a per-deployment basis, depending on the site's machines. Automatic braking is not universally deployed: some sites have found that throughput impact makes it impractical, and the configuration should match the site's operational requirements.

Does OSHA require pedestrian detection systems?

No. OSHA does not mandate a specific detection technology. Operator training is required to cover pedestrian traffic (29 CFR 1910.178(l)(3)(ii)), and the General Duty Clause requires employers to address recognized hazards. Detection systems are an engineering control that supports compliance, but they are not individually required by name. OSHA also publishes no numeric safe-distance rule for pedestrians around forklifts. Figures such as "stay 10 feet back" are employer and insurer conventions.

Will pedestrians need to wear anything?

Not for camera-based systems. Both onboard and site-level camera systems detect people by sight, with no wearable required. Only tag-based proximity systems, such as RFID or UWB, require pedestrians to carry or wear a device. The no-wearable characteristic of camera systems is particularly relevant for sites with high contractor, visitor, or agency worker traffic.

How do I measure whether pedestrian detection is working?

Track the number, location, and timing of logged near-miss events over time. A system configured correctly will first reveal the true frequency of close calls, which is almost always higher than the written log suggests before monitoring begins. Targeted interventions based on that data, including zone changes, traffic rule updates, and coaching, should then produce a measurable reduction in event frequency. Ailytics customers in logistics and warehousing report up to 80 percent reductions in safety non-compliance through this cycle.

Keep reading

Find out what your cameras could catch

Tell us about your sites and cameras. We will show you which detections fit and how alerts reach your team.

Request a demo