A camera can provide a useful overview without providing the view needed by a particular analytics application. Before moving it or adding another camera, define the event and identify what must remain visible when that event happens.
Placement is one part of the design. Stream settings, image quality, processing and the application's supported conditions also matter. Do not assume that a new angle alone will solve every detection problem.
Start with the event and target area
Write one observable requirement: for example, a person crossing a defined boundary after closing. Mark the intended area, direction of movement and operating hours. List normal activity that should not produce the same response.
A face-recognition or number-plate application has different capture requirements from general person or vehicle detection. Do not apply one mounting rule to all three. Establish the manufacturer's requirements and the proposed integration before choosing a final position.
Where identity is involved, agree the permitted use, enrolment and review process with the responsible team. A detected person is not automatically an identified person.
Inspect visibility through the whole event
Check the approach, target area and departure rather than a single attractive still frame. Look for doors, columns, parked vehicles, stacks or people that obstruct the relevant view. Include normal operating changes between shifts.
Record whether the required features remain visible for long enough for the proposed application. Axis's object analytics documentation describes application-specific considerations including mounting, object visibility and lighting. Those settings are not universal instructions for another product.
A plan that works with an empty floor may need revision when the factory or warehouse is operating normally. Keep the untested conditions explicit.
Check height, angle and detail together
There is no single correct mounting height for every event or camera. The subject, lens, distance and application requirements need to be considered together. Also allow for physical protection, authorised access and maintenance.
A steeper angle may hide features needed for one task while serving another overview task. A narrower view may improve detail at the target while leaving another route outside the frame. Record the trade-off rather than declaring every high camera unsuitable.
Use the pixel-density guide to understand available detail. Human DORI planning benchmarks should not be substituted for the analytics supplier's own capture and testing requirements.
Observe the lighting that matters
Review the view during relevant daylight, shutter positions and night conditions. Note backlighting, reflected light, headlights and obscured or dirty covers. Look at the actual recording or processing stream, not only a high-quality configuration preview.
Do not treat an advertised infrared range as proof of recognition or event performance at that distance. Ask what the range describes, then test the required task. A lighting change should be designed for the site and camera rather than promised as a cheap universal fix.
Test a proposed position before committing
Agree a safe temporary evaluation with the installer and site owner. Keep the existing security arrangement operational as required while assessing the alternative.
| Record | Why it is needed |
|---|---|
| Camera, lens and settings | Identifies what was actually tested |
| Position and target distance | Allows the view to be reproduced |
| Event and negative examples | Defines which results should and should not occur |
| Day/night conditions | Shows the scope of the evidence |
| Misses, unwanted alerts and delay | Exposes limitations beyond the best clip |
| Accepted change or next test | Connects the result to a decision |
Use the pilot worksheet for the event record. A passed demonstration at one position does not approve every camera on the site.
Keep the commissioned view under review
Store the agreed view and configuration with the handover record. Reassess it after racking changes, new signs, moved machinery, seasonal storage or a camera adjustment. A small physical change can alter the scene the test originally covered.
For an AI Alerto by PGAK assessment, prepare the event, location and existing model details. The compatibility checklist helps organise that information without sharing passwords or unrestricted camera access. The objective is a documented, testable view for the intended task.