Define the question before choosing the view
Begin with the proposition a review must establish. Did the whole ball cross a boundary? Where was an eligible body part when the ball was played? Was contact visible, and what preceded it? These questions require different evidence. A camera selected because it looks dramatic may be poorly placed to answer the actual question. Record the incident and the relevant rule before choosing frames or adding an overlay.
Keep the original decision alongside the review question. In some procedures the evidence must justify changing that decision; in others a designated system supplies an authoritative factual indication within a narrow scope. Do not transfer the standard from one sport or competition to another. The IFAB’s football protocols, for example, are not the operating instructions for an International Tennis Federation competition.
Establish a common geometry
A camera converts a three-dimensional scene into a two-dimensional image. Perspective can make separated objects appear aligned, while lens distortion changes apparent shapes near the edges. Calibration relates image coordinates to known features of the playing area and, where required, to a shared three-dimensional coordinate system. A reliable reconstruction depends on that relationship remaining valid after installation.
Camera movement, a changed zoom setting or a shifted mounting point can disturb the relationship. Multiple viewpoints can reduce ambiguity, but only when their calibration and synchronisation are appropriate. A broadcast line drawn on a convenient view is not automatically a validated measurement. Ask whether the graphic is explanatory television production or part of the approved officiating workflow.
Checkpoint: preserve the source
Retain the relevant original views and identify which images were cropped, slowed, enhanced or reconstructed. An explanatory graphic should never quietly replace the underlying record.
Locate the relevant moment
Position matters only in relation to time. A frame rate gives discrete samples rather than an uninterrupted account, and exposure may blur movement within each sample. Selecting the first frame that appears to show contact can still leave uncertainty about the actual contact time. Rolling-shutter effects can also mean that different image rows were not captured at precisely the same instant.
Where several feeds are combined, their timestamps must refer to a sufficiently consistent clock. An offset can make one view appear to contradict another even when both recorded the event correctly. Systems may use dedicated timing infrastructure or synchronisation methods such as those described by IEEE 1588. The presence of a standard, however, does not prove the complete installation meets its operational timing requirements.
Recognise reconstruction as an estimate
Computer vision can detect a ball, track players or estimate body landmarks. Occlusion remains central: a leg behind another player is not directly visible just because software draws its likely location. Training data, camera angle, clothing, lighting and unusual postures can affect performance. A rendered skeleton should therefore be understood as a model output, not a photograph of an otherwise hidden joint.
Human operators may inspect or correct particular outputs within an approved workflow. That step should be traceable, especially when a corrected point changes a close positional result. Do not assume every installation uses the same amount of automation. FIFA’s descriptions of semi-automated offside technology concern particular systems and processes, not a universal promise about all computer-vision products.
| Observation | Inference or decision |
|---|---|
| A ball is visible in several calibrated views. | A reconstruction estimates its position between sampled moments. |
| An image shows contact between players. | The referee judges whether the contact constitutes an offence. |
| A body landmark is obscured. | A model predicts its location; direct visibility has not been restored. |
Apply the law after measuring
A geometric finding can settle a narrow issue without settling the whole incident. In football, positional offside evidence does not alone decide every question of involvement in active play. Likewise, a replay showing contact does not independently decide whether a challenge was careless, reckless or used excessive force. Those are applications of the laws, informed by context and the authorised review procedure.
Use slow motion to inspect a sequence, but return to normal speed where speed and force affect interpretation. Repetition can make a marginal touch feel more significant without adding information. The final account should identify what the cameras established, what remained estimated and which judgement the official made. If the decisive area stayed hidden, say so and apply the competition’s fallback rather than manufacturing certainty.
Primary reading
Consult the current competition documents for operating requirements. These organisations do not endorse this website.