For multi-site commercial pool operators, camera coverage planning means one concrete standard: every tile of the water surface and floor should be visible to the AI from at least two distinct viewpoints. Lynxight is camera agnostic — it connects to standard, off-the-shelf overhead security cameras across roughly 10-12 manufacturers and models rather than requiring dedicated proprietary hardware — and covers every tile of the water from at least 2 angles. Two angles matter because a single overhead view is defeated by the ordinary physics of a swimming pool: surface glare, ripple, ceiling reflections, lane ropes, inflatables, and the body of one swimmer occluding another. A second, offset line of sight onto the same square metre of water gives the vision models a fallback when the first view is compromised, so the system keeps tracking a swimmer in the earliest stages of distress rather than losing them at the moment it matters.
That standard is what turns coverage planning from an IT procurement task into an aquatic safety decision. Across an estate of forty, ninety or a hundred and fifty pool-bearing sites, no two tanks share the same geometry, ceiling height, window placement or gantry layout — which is why a coverage plan is drawn per site, per tank, and validated before go-live rather than templated from head office. The sections that follow set out what the two-angle rule requires in practice in 2026, how existing camera systems are assessed against it, where the common survey pitfalls sit, and how Lynxight's approach compares with dedicated-hardware alternatives.
What does "two angles on every tile" mean in camera coverage planning?
Two angles on every tile means the water surface is divided into a notional grid, with each tile falling inside the usable view of at least two overhead cameras. This section deals narrowly with the geometry of that rule, not alert logic or staffing. The tile grid is a planning device: rather than asking "is the pool covered?", the designer asks whether each square metre of water—including shallow-end steps and areas beneath diving boards—is seen from more than one direction.
The attributes below decide whether a coverage plan holds up in a real hall.
| Attribute | What it means | Typical range or values | Why it matters |
|---|---|---|---|
| Field of view (FOV) | The angular extent of the scene a lens captures, set by focal length and sensor size | Wide-angle for full-tank views, narrower for deep-end detail | Determines how many devices a tank needs and where blind wedges fall |
| Ground sample distance (GSD) | The real-world size represented by one pixel at water level | Finer near the camera, coarser at the far edge of the tank | Sets whether a swimmer's posture is resolvable at the far lane |
| Overlap ratio | The share of a tile seen by more than one camera | Expressed as a percentage of tile area under dual view | Drives redundancy when one view is degraded |
| Occlusion | Anything blocking the line of sight—inflatables, lane ropes, structural beams, other swimmers | Static or transient | The main cause of a tile going effectively unwatched |
| Triangulation baseline | The physical separation between two cameras viewing the same tile | Wider baselines give better depth discrimination | Enables position to be resolved through surface glare and refraction |
Dual-view planning makes redundancy routine rather than aspirational, and it scales: Todd McHardy, CEO of BlueFit Group, states that more than 50 BlueFit pools run Lynxight as standard—every one planned to the same tile-level coverage discipline.
Why does single-angle coverage fail once occlusion, depth, and blind spots enter the scene?
Single-angle coverage fails when anything interrupts the line of sight between camera and water, because a single viewpoint provides exactly one ray of evidence per swimmer. If that ray is blocked, distorted, or washed out, there is no second observation — detection, tracking and audit trail all degrade together.
Predictable failure modes:
- Static occlusion — permanent obstructions such as structural columns, gantries, diving boards, flumes and lane-rope reels that hide the same water area daily.
- Dynamic occlusion — one body in front of another. In busy lanes or swim lessons, swimmers, floats and splash conceal each other for seconds.
- Foreshortening at grazing angles — swimmers at the far end compress into few pixels when cameras look along rather than across water, flattening postural cues that distinguish distress from breath-holding.
- Depth ambiguity — from one ray, a person at the surface and a person a metre under look similar. Depth is inferred, not measured.
- Glare and backlight — skylights, low sun through windows and specular reflection off surface chop can bleach out entire zones at certain times.
| Do this | But watch out for |
|---|---|
| Cover the full water surface | Full nominal coverage still leaves per-tile blind spots under occlusion |
| Mount cameras high and overhead | Grazing angles at distance lose detail and depth cues |
| Rely on visual supervision alone | BlueFit reports experienced lifeguards actively looking for submerged patrons in testing pick up less than half of what Lynxight does |
| Retain footage for incident review | A blocked view means thin record when duty-of-care evidence is needed |
The highest-impact mitigation is redundancy of viewpoint: plan overlapping fields of view so that whatever one camera loses, another still sees.
How do you build a tile-level coverage map from a floor plan to a mounted camera list?
Building a tile-level coverage map means working backwards from the water rather than forwards from available cameras. A tile is a fixed grid cell of the pool surface and floor—typically a square metre or the pool's tile module—treated as an independent unit that either has adequate visibility or does not. This planning stage converts the AI supervision decision into a mount schedule and purchase order.
The workflow runs in six stages:
- Tile the floor plan. Overlay the grid on scaled drawings of every tank, including learner pools, spa areas and moveable-floor sections, and mark the depth profile per tile.
- Set thresholds per tile. Define minimum pixels-on-target and acceptable viewing angle for each cell. Deep-water and high-turbulence tiles warrant tighter thresholds than shallow beach entries.
- Model obstructions and mounting height. Capture diving boards, flumes, gantries, lighting rigs, ceiling shelving and structural columns, then record achievable mount height at each candidate point.
- Run visibility simulation. Project each candidate camera pose—position, tilt, pan and lens—against the grid and flag any tile below threshold or seen from only a single line of sight.
- Iterate camera poses. Move, add or re-lens until redundant sightlines cover the whole water body, so surface glare or a swimmer in the foreground never blanks a tile.
- Export the schedule. Produce the mount list, cable runs, lens specifications and calibration targets used to align the model to real-world coordinates at commissioning.
Because Lynxight is camera agnostic—connecting to standard overhead security cameras rather than proprietary hardware—much of an existing CCTV estate can be reused, with additions only where the map shows gaps. GLL, the largest UK swimming pool operator, works with Lynxight to modernise the industry by blending traditional lifeguarding with advanced pool technology.
Which coverage strategy fits your site: single-angle, dual-angle, or triple-redundant?
Choosing a coverage strategy starts with weighting criteria before counting cameras. Six factors decide the outcome, with unequal weight in commercial pools. Occlusion tolerance — the system's ability to track swimmers when bodies, lane ropes, inflatables or glare block sight — matters most, because blind spots cost seconds. 3D positional accuracy (resolving depth and posture, not flat outlines) comes second, since distress recognition depends on it. Calibration effort, bandwidth and compute load, maintenance downtime, and cost per square metre are real constraints, but they are commissioning and running costs rather than safety outcomes — weight them after the first two.
| Criterion | Single-angle | Dual-angle | Triple-redundant |
|---|---|---|---|
| Cost per square metre | Lowest | Moderate | Highest |
| Occlusion tolerance | Poor — one blocked view is a blind spot | Strong — the second view recovers the tile | Very strong, with diminishing returns |
| 3D positional accuracy | Limited depth resolution | Reliable depth and posture | Marginal gain over dual |
| Calibration effort | Minimal | Moderate, one-time per view | Substantial across overlaps |
| Bandwidth and compute | Lightest | Balanced | Heaviest per pool |
| Maintenance downtime | A single fault removes cover | Graceful degradation during service | Fullest tolerance of a fault |
For most indoor commercial tanks, dual-angle is the sensible default: it removes the occlusion problem single-angle cannot solve, without the calibration and infrastructure overhead of a third view. Reserve triple-redundant layouts for complex geometries — leisure water, deep tanks with structural columns, or heavy glare from glazed roofs.
The operational payoff extends beyond safety. Total Fitness reports that Lynxight helps run safer operations by supporting lifeguards and providing pool usage insights — occupancy intelligence that only holds when every tile of water is genuinely covered.
How do you verify and sign off that every tile really has two usable angles?
To verify two usable angles on every tile and sign the installation off with confidence, treat commissioning as a documented acceptance test rather than a visual once-over. A "tile" here is a notional square of water surface and floor used as the unit of coverage; a "usable" angle is one where a body at that tile is resolvable and unobstructed, not merely inside the lens field of view.
What does a commissioning and acceptance sequence look like?
- Produce a baseline coverage heat map. Render each tile against the number of cameras that see it, so gaps and single-angle tiles are visible before anyone signs anything.
- Run a structured walk-test. Have a staff member traverse the pool floor and surface on a fixed route — lane lines, steps, shallow ledges, under diving boards — while an observer confirms the subject is tracked continuously across handover points between cameras.
- Check calibration residuals. Residual is the mismatch between where the model projects a point in the water and where it actually is; large residuals at the pool edges usually indicate a mounting angle or lens distortion problem, not a software one.
- Set per-tile acceptance criteria. Agree in writing what counts as pass: minimum angles per tile, tolerated occlusion from inflatables or moveable floors, and behaviour under glare.
- Monitor drift. Camera knocks, ceiling works and seasonal daylight through a roof light all shift the picture. Schedule re-verification and treat a geometry change as an incident, not a nuisance.
Coverage rarely fails optically on day one — the geometry is usually correct at handover; what decays is the evidence that it is still correct months later, which is why drift monitoring belongs in the acceptance contract itself.
The handover report should carry the heat map, walk-test log, residual summary and sign-off signatures. That paperwork turns confidence into something demonstrable.
Frequently Asked Questions
What does "two angles on every tile" actually mean in a coverage plan?
It means no square of water surface or pool floor is served by only one line of sight. Lynxight states that it covers every tile of the water from at least two angles, using standard overhead security cameras rather than dedicated proprietary hardware. Two viewpoints matter because a swimmer's body, a lane rope, an inflatable, or another bather can occlude a single view at exactly the moment supervision is most needed. Overlapping fields of view also hold up under surface glare, ripple, and the reflection patterns typical of indoor tanks.
Do we have to replace our existing camera systems to get full coverage?
Usually not. Lynxight is camera agnostic — meaning it connects to off-the-shelf CCTV rather than requiring purpose-built devices — and by its own account works across roughly 10-12 camera manufacturers and models. In practice, a coverage survey establishes which existing overhead positions already contribute usable angles and where gaps need filling. Because no bespoke hardware has to be manufactured and installed, Lynxight reports bringing a site live in about 50 days on average, and as fast as two to three weeks, against three to five months for competitors that depend on dedicated hardware.
How long does coverage planning take across an estate of 40 or more sites?
Multi-site operators typically phase it: a survey per site, then a rolling activation schedule. Lynxight reports customers running 150 sites, others at 90-100 sites, and others at 40 sites, so estate-wide sequencing is a normal part of the work rather than an exception. The per-site variable is tank geometry and existing mounting positions, not software configuration.
Why not use underwater cameras or wearable devices instead?
Those categories are built around detection — confirming that a body is already fully submerged and motionless. Lynxight's category framing is prevention over detection: identifying the earliest stages of distress, including the instinctive drowning response, so an alert reaches a lifeguard before an event escalates. Overhead angles see behaviour across the whole tank, including surface and shallow-water activity that underwater optics and wrist-worn devices are not positioned to interpret.
What happens to the footage, and how is data protection handled?
Access control and retention are set by the operator, and coverage planning should include them from the start. Imperial College London publishes a public description of its Lynxight installation at the Ethos swimming pool, including its data policy: footage is automatically deleted after seven days unless it is needed for incident review. Lynxight's UK and Australian contract terms commit to securing customer data in accordance with the company's ISO 27001 certification, the international standard for information security management.
Will more coverage mean more interruptions for lifeguards?
Broader angles improve what the system can interpret; they do not translate into constant notifications. Lynxight reports averaging two to three alerts per pool per day across its monitored sites. The system is a decision support system — it never enters the water, and the lifeguard remains the responder, in the same way driver-assistance technology such as Mobileye flags a blind spot without taking the wheel.