How to measure forklift-pedestrian exposure with existing cameras

The useful question is not whether a camera can see a forklift. It is whether the view can support a repeatable measurement of a defined forklift-pedestrian condition.

This guide separates event detection from measurement. The goal is to define the visible condition, qualify the camera, review distinct episodes, normalize by usable observation time where appropriate, find recurrence and recheck the same condition after the site acts.

What is forklift-pedestrian exposure?

For camera-based measurement, forklift-pedestrian exposure should be defined as an observable relationship in a specific operating area. Examples include a pedestrian entering a forklift operating zone, a forklift entering a pedestrian-protected area, a forklift and pedestrian remaining within a site-configured separation, or repeated interaction at one crossing.

That wording matters. A camera can observe where people and equipment are in relation to a defined area or to one another. It cannot infer whether a person intended to take a risk, whether a movement was authorized, or whether the site has satisfied every element of a safety requirement.

Start with the visible condition, not a label about worker behaviour. A bounded definition makes the result reviewable and gives the site something concrete to measure again after it changes the operation.

Can existing CCTV or security cameras be used?

Often they can, but not every camera is a measurement camera. A useful view needs to show enough of the relevant travel path, crossing or zone to support the selected condition. A camera that is excellent for general security may still be unsuitable if forklifts disappear behind racks, the pedestrian route is outside the frame, or the geometry makes the relationship ambiguous.

A camera-fit review should therefore answer a few basic questions before measurement begins:

• Are the forklift and pedestrian travel paths visible where the condition matters?
• Can the relevant zone, crossing or boundary be represented clearly?
• Do occlusion, distance and lighting allow reliable observation during the intended operating windows?
• Will the same view remain usable if the site changes a barrier, route or layout?

If the view cannot support the question, the correct result is not a lower-confidence safety claim. The correct result is to change the view, narrow the question or withhold the measurement.

Define the condition before collecting the evidence

Broad goals such as "improve forklift safety" are useful program objectives, but they are not measurement definitions. The operational condition needs a place, an observable relationship and, where relevant, a site-defined threshold or time window.

For example, a site might choose one shared cross-aisle and ask whether reviewed forklift-pedestrian episodes keep recurring there during a particular operating window. Another site may focus on pedestrian entry into a forklift-only zone. The goal is not to encode every safety rule at once. It is to produce one result that a safety and operations team can inspect, understand and act on.

Turn camera observations into reviewed operating episodes

Continuous video creates many frames, but the business question is not how many frames contained a person and a forklift. The useful unit is a distinct operating episode that represents the configured condition and has enough supporting evidence for review.

That distinction prevents a short interaction from being counted hundreds of times simply because it lasted for many frames. It also creates a record that can retain the relevant clip, timestamp, camera, zone and review status together.

The result should describe a set of reviewed operating episodes, not a stream of machine signals.

Look for recurrence by place and time

One reviewed episode may deserve immediate attention. A recurring concentration answers a different question: where and when does the condition keep returning?

Organizing reviewed episodes by area and operating window can reveal whether the condition is broadly distributed or concentrated around a particular crossing, doorway, staging area or shift transition. That pattern gives Operations and EHS a more specific place to investigate than a generic instruction to "be more careful around forklifts."

Counts also need an observation basis. If one period contains more usable camera time than another, raw counts can mislead. A defensible result states the usable observation time and uses a comparable basis when presenting rates or before-and-after change.

Measure the same condition again after the site acts

The value of a recurring pattern is that it can support a bounded operating decision. The site may change a route, add physical separation, alter staging, change timing or choose another control based on its hazard assessment.

After implementation, the same camera and measurement definition can be used again if the observation conditions remain comparable. The follow-up can then report whether the measured condition increased, decreased or remained similar.

A before-and-after comparison is not automatically a causation claim. Work volume, staffing, schedules, visibility and other operating conditions may also change. The result should state the observation basis and the limits of what the comparison can establish.

What camera evidence cannot establish

Camera evidence can be useful precisely because its boundary is explicit. It can establish what was visible in the qualified view under the configured condition. It should not silently expand into claims the camera cannot support.

• It does not identify a worker or score an individual's safety performance.
• It does not establish intent, permission or authorization from visual evidence alone.
• It does not replace the employer's hazard assessment or control decision.
• It does not prove that a measured change was caused only by the intervention.
• It does not describe activity outside the usable camera coverage.

See a complete forklift-pedestrian measurement example

Edgentik's public sample shows 34 reviewed forklift-pedestrian episodes across 96 usable hours from recorded demonstration footage, followed by a bounded recurring-pattern finding and an illustrative matched-basis recheck. It is not a customer result and the follow-up is descriptive rather than causal.

Related analysis: why recurring safety patterns require a different standard from individual alerts, and how recurring conditions can support industrial safety leading indicators.