Skip to content
Safety Tech Review
Menu

Near misses

Near-Miss Detection Explained: What AI Video Counts as a Close Call

How AI video defines a near miss, how proximity and time-to-collision rules work, why counts mislead, and how to write a near-miss definition your site can trust.

By · Updated · 10 min read · 10 sources

Reviewed by LIPAI WANG, Editor, on against the cited sources. How we review

A near miss is an event that could have hurt someone but didn't. AI video safety systems detect a narrow slice of these, mostly catching people getting too close to moving vehicles, loads, or machines. To do this, the software applies rules for distance, speed, zones, and timing to whatever the cameras see. Those rules are simply a proxy for risk. So the number on a dashboard depends as much on how the site defines a near miss as on what actually happened. This article explains the standard definitions, how camera systems turn them into rules, where the counts mislead, and how to write a definition you can defend.

What counts as a near miss?

Safety regulators define a near miss by what could have happened. HSE's guidance on investigating accidents and incidents (HSG245) defines a near miss as "an event that, while not causing harm, has the potential to cause injury or ill health," and for the purposes of that guidance includes dangerous occurrences in the term [2]. HSG245 separates this from an "undesired circumstance," a set of conditions or circumstances with the potential to cause injury or ill health, such as untrained nurses handling heavy patients [2]. The same guidance points out that it is often luck that decides whether an undesired circumstance becomes a near miss or an accident [2].

OSHA's template near-miss reporting policy takes a broader view. It describes a near miss as an opportunity to improve safety based on a condition or incident with potential for more serious consequences, and lists unsafe conditions, unsafe behavior and minor incidents that could have been worse [1]. Many US employers use a policy like this. But OSHA does not generally require near misses to be reported to the agency.

In Great Britain, some events in this family are legally reportable. RIDDOR lists dangerous occurrences, which HSE describes as certain incidents with a high potential to cause death or serious injury, such as the collapse or overturning of lifting equipment or contact with overhead power lines [3]. These must be reported whether or not anyone was hurt, and whether or not a camera recorded them.

All three definitions share one feature that matters for technology: they rest on judgment about potential. A person deciding "that could have killed someone" is weighing speed, mass, distance, the people involved, and what might have happened next. A camera system has to approximate that judgment using basic measurements.

How does AI video decide something was a near miss?

Camera-based systems detect objects (people, forklifts, trucks, loads), track them across frames, and apply rules. A near miss is simply whatever the rules say it is. Intenseye, for example, lists detections such as "person in an active forklift path," "converging paths with no sightline" and a worker beneath a lifted load [6]. Protex AI says it recognizes forklifts, pallet jacks, automated vehicles and other vehicles, monitors pedestrian movement alongside them to analyze human-machine interactions, and detects near misses [7].

The common rule types are shown below.

Rule type What the system measures Typical example What can go wrong
Proximity breach Distance between a person and a moving vehicle or machine Pedestrian within a set distance of a moving forklift Distance estimates depend on camera calibration and angle
Path conflict Predicted paths of a person and a vehicle crossing within a time window Forklift and walker approaching a blind corner Predictions are short range and assume constant speed
Zone intrusion during operation A person inside a drawn zone while equipment is active Person inside a robot cell or under a crane load Depends on knowing when the equipment is live
Evasive action Sudden braking, swerving or a person stepping back quickly Driver brakes hard as a pedestrian steps out Hard to separate from normal maneuvering
Line of fire A person between a hazard and its path of release Worker standing behind a reversing truck Requires knowing the direction of travel

Each rule has parameters: a distance, a speed floor, a time window, a minimum duration, an active-hours schedule. A site that sets a 3 meter proximity threshold will log far more events than one that sets 1 meter, from the same footage, so any count is only comparable with counts made under the same settings.

What road safety research says about using conflicts as a proxy

Traffic engineers faced the exact same problem decades ago. Crashes are rare. Waiting for them is a slow way to judge whether an intersection is dangerous. The traffic conflict technique, in which observers count strong braking and evasive maneuvers, formally began with studies at General Motors Research Laboratories in the late 1960s [5].

The US Federal Highway Administration's 2008 report on its Surrogate Safety Assessment Model (SSAM) shows how the idea became automated. SSAM classifies vehicle interactions as conflicts using two measures. One is time to collision (TTC), the time remaining before two vehicles would collide if they kept their current speed and path. The other is post-encroachment time (PET), the gap between one vehicle leaving a point and another arriving at it [5]. SSAM uses a default TTC threshold of 1.5 seconds, based on earlier research, and lets analysts change it [5].

The same report is candid about the limits. It says the conflict method has shown some correlation with crashes, but that there is still debate about the connection between conflict measures and crash predictions [5]. Workplace systems that borrow TTC-style logic for forklifts and pedestrians inherit that debate. And they do so with less validation data behind them than road safety has.

Is a camera near miss the same as a reported near miss?

No, and treating them as one series will distort trends. The two capture different events and fail in different ways.

Feature Manual near-miss reports Camera-detected near misses
Who decides The worker or supervisor who saw it A rule set configured by the site and vendor
Coverage Anywhere people work, any hazard type Only areas in camera view, mostly visual and movement hazards
Volume Usually low, limited by willingness to report Often high, limited by thresholds and review capacity
Context Can include intent, cause and what happened next Usually a short clip and metadata; Protex AI, for example, says it uploads a 10-second event clip [8]
Main bias Under-reporting, especially of embarrassing or routine events Threshold choice, camera placement and detection errors
Best use Learning about causes, culture and hazards cameras cannot see Finding where and when people and vehicles mix

Keep the two as separate series. A site that adds camera events to its manual near-miss total will see an apparent jump in near misses the month the system goes live. That jump says nothing about actual risk.

Why near-miss counts can mislead

Thresholds and definitions drift

Vendors and sites often re-tune rules during pilots to cut nuisance alerts. But every change resets the baseline. If the proximity distance moves from 3 meters to 2 meters in month two, a drop in events from month two onward is just a change in measurement. Log every rule change with a date so trends can be read correctly.

Exposure changes

More forklift movements, more shifts, or a seasonal peak will naturally produce more conflicts at the exact same level of risk per movement. Report near misses per hour of operation, per thousand vehicle movements, or per camera-hour alongside the raw count.

Falling counts are not proof of safety

Vendor case studies often report large drops in detected near misses. Protex AI's website cites an "80% reduction in near-miss events within 54 days of deployment" at a North American automotive manufacturer, and a 97% decrease at a food distribution company [7]. These are vendor-reported results and have not been independently verified. A fall in detected events can absolutely come from safer behavior, which is the goal. But it can also come from re-tuned thresholds, a camera knocked out of alignment, workers avoiding the monitored area, or just reduced traffic. Before accepting a reduction, ask what changed in the rules and the coverage over the same period.

The pyramid does not hold as well as the poster suggests

The appeal of near-miss counts rests on the safety pyramid. Frank Bird's work, based on more than 1.7 million accidents reported by 297 companies and disseminated in the 1970s, led to a ratio usually quoted as 1-10-30-600 [4]. Later studies have questioned whether such ratios are actually stable or predictive. One study of US mines found no significant difference in the probability of a fatal event the following year between mines with 16 or fewer total lost or restricted workdays and those with none, and research in Chile found the ratios vary with the base incident rate [4]. A camera system that drives down minor proximity events might not touch the precursors of a fatality. Those precursors often involve different tasks, such as maintenance, lifting, or work on unguarded machinery.

What camera near-miss detection cannot see

Cameras can only flag what is in view and visible. That rules out most electrical near misses, chemical releases below a visible threshold, pressure system faults, and many ergonomic close calls. It also rules out events behind racking, inside vehicles, or in areas without coverage. RIDDOR dangerous occurrences, such as a lifting equipment collapse, still need reporting through your normal process even if a camera captured the clip [3].

Proximity rules also struggle to judge intent and control. A banksman standing close to a reversing truck is doing their job; a visitor in the same spot is a near miss. Some systems handle this with zones and schedules. Others try excluding people wearing particular clothing. Each workaround brings its own error rate.

How to write a near-miss definition your site can use

Start from the hazards most likely to cause serious harm on your site. Then write rules that a reviewer can apply to a clip and consistently get the same answer as a colleague.

  1. List the interactions that matter: forklift and pedestrian, truck and pedestrian at the dock, person under a suspended load, person inside a machine envelope. OSHA's pedestrian traffic guidance for powered industrial trucks starts from separating forklift traffic from pedestrians where possible, which tells you where to look first [9].
  2. For each interaction, set measurable conditions: distance, vehicle moving or stationary, speed floor, minimum duration, zone and active hours.
  3. Assign severity tiers. A pedestrian 0.5 meters from a forklift moving at speed is a different event from one at 2 meters beside a slow vehicle. Tiering lets you prioritize review and keeps the headline number from being dominated by low-risk events.
  4. Define what the reviewer confirms: the objects were correctly detected, the conditions were met, and the event meets the site definition. Record false positives by cause.
  5. Fix the definition for a set period, at least a quarter, and change it only through a logged decision.
  6. Normalize the count by exposure and publish the definition next to the chart.

A short definition table might look like this.

Event Conditions Tier Reviewer action
Pedestrian and moving forklift, close Under 1 m, forklift moving, 0.5 s or longer High Review every event within one shift
Pedestrian and moving forklift, near 1 to 2 m, forklift moving Medium Review weekly sample
Person in crane exclusion zone Inside zone while load suspended High Review every event
Pedestrian outside walkway in vehicle aisle Any time during operating hours Low Trend only, sample monthly

The numbers above are just illustrations of the format. Your thresholds should come from your risk assessment, your vehicle speeds, and the measurement accuracy your vendor demonstrates on your own cameras.

Watch the alert load

A system that produces hundreds of low-value alerts just teaches people to ignore all of them. A 2006 NIOSH evaluation of a radar proximity warning system on haul trucks found it detected people and vehicles reliably. But frequent alarms from objects that posed no immediate danger were a real problem, and the author recommended pairing sensors with cameras so operators could check the cause [10]. Send only high-tier events as live alerts. Route the rest directly to trend reports.

Using near-miss data without punishing workers

Camera near misses often show a person in the wrong place. The tempting response is discipline. HSG245 makes the case for investigating near misses because doing so is as useful as investigating accidents and much easier, and because witnesses are more likely to be helpful and tell the truth when no one has been harmed [2]. If camera events lead mainly to individual warnings, workers will simply learn to avoid the cameras. Then the useful signal about layout, scheduling, and traffic design is lost entirely. Instead, use clusters of events to ask why people cross where they do, whether walkways actually match real routes, and whether vehicle and pedestrian flows can be separated by time or space.

Summary

A near miss is defined by its potential for harm, and that judgment has to be turned into rules before a camera can count one. Most AI-detected near misses are proximity, path-conflict or zone events involving people and vehicles or loads, and the count depends heavily on thresholds, camera coverage and traffic volume. Road safety has used conflict counts as a proxy for crashes since the 1960s and still debates how well they predict harm. Treat camera near misses as a separate series from manual reports, write and freeze a site definition with severity tiers, normalize by exposure, audit a sample, and read vendor claims of large reductions as vendor claims until you have checked what changed in the rules.

Frequently asked questions

+Do I have to report AI-detected near misses to OSHA or HSE?

No general duty exists to report near misses to OSHA, and an AI alert does not change that. In Great Britain, RIDDOR requires reporting of listed dangerous occurrences, such as the collapse of lifting equipment, whether or not a camera saw them. Treat camera events as internal data, and keep your statutory reporting process separate.

+Is a higher near-miss count good or bad?

Neither, on its own. With manual reporting, a rise often reflects a healthier reporting culture. With camera detection, a rise may reflect more traffic, more camera coverage or a looser threshold. Look at the rate per hour of exposure and at the severity mix before drawing a conclusion.

+Can AI video predict serious injuries from near misses?

Not reliably on current evidence. Near-miss data points to where people and hazards meet, which is useful for prioritizing controls. Studies of the Heinrich and Bird ratios have found the link between minor events and fatal ones is weaker and less stable than the pyramid suggests.

+How many near-miss events should a reviewer check?

During a pilot, check every high-severity event and a random sample of the rest each week, and record whether each one meets the site definition. Once precision is stable, a smaller sample is usually enough, but keep sampling so drift is caught.

Sources

  1. [1]Template for Near Miss Reporting Policy (OSHA)
  2. [2]Investigating accidents and incidents, HSG245 (HSE)
  3. [3]RIDDOR: Dangerous occurrences (HSE)
  4. [4]The Heinrich/Bird safety pyramid (Risk Engineering)
  5. [5]Surrogate Safety Assessment Model and Validation: Final Report, FHWA-HRT-08-051 (FHWA, 2008)
  6. [6]Real-Time AI Safety Solutions for SIF Prevention (Intenseye)
  7. [7]Protex AI home page (customer results)
  8. [8]Privacy at Protex AI
  9. [9]Powered Industrial Trucks eTool: Pedestrian Traffic (OSHA)
  10. [10]Ruff, Evaluation of a radar-based proximity warning system for off-highway dump trucks, Accident Analysis and Prevention (2006)

Get the free Safety Tech Buyer's Toolkit

A 15-page PDF: the public evidence checklist, a weighted vendor scorecard, 40 RFP questions and pilot acceptance criteria. Subscribe and download it straight away; then one email a month when we publish or update guidance. No vendor promotions. Unsubscribe any time.