Tune image-health alerts with human validation

Bottom line: an image-health alarm can indicate that a view changed; it cannot decide whether the change is a fault, a deliberate maintenance action, normal scene behavior, or a security event. Treat it as a triage signal backed by an accountable response.

Source fact: the analytic detects defined image conditions with limits

The AXIS Image Health Analytics User Manual documents detection of blocked, redirected, blurred, and underexposed images. It provides sensitivity and validation-period controls. The manual also says the analytic cannot distinguish intentional from accidental camera movement and describes behavior for PTZ cameras, including suspension during pan, tilt, or zoom activity.

These characteristics create a tuning problem. A short occlusion may be normal at a loading dock, while a small shift can be serious at a narrow doorway. One global delay or sensitivity value will not express both operational realities.

Source boundary and applicability

The manual applies to compatible Axis devices and documented software versions. It does not guarantee detection of every obstruction, focus problem, attack, illumination loss, or view change. It also does not replace camera-heartbeat, recording, storage, time, or network monitoring. Confirm compatibility and supported behavior on the exact device.

Applicability questions

• Which health condition matters for this camera’s security task?
• What normal movement, cleaning, weather, doors, vehicles, or lighting changes could create alerts?
• How long may the view be degraded before response is required?
• Does the PTZ have tours or operator movements that suspend analysis?
• Who verifies the image, creates a ticket, and escalates suspected tampering?

DSE recommendation: test degradations and the response chain

The following steps are DSE recommendations based on the cited source.

Define the expected view and response class for every critical camera. In an approved window, introduce reversible examples of each supported condition: partly and fully block the view, shift the camera within and beyond tolerance, defocus it, and reduce illumination. Do not damage equipment or interfere with active security coverage. Measure detection time and recovery behavior, then tune sensitivity and validation period from those results.

Route alarms with camera identity, current image, condition, timestamp, and runbook. Require an operator or technician to classify the event and confirm restored coverage. Suppress alerts only through documented maintenance mode with an owner and expiration.

Verification and evidence

Retain the baseline image, model and version, analytic configuration, test matrix, alert and recovery timestamps, notification evidence, ticket disposition, false-positive observations, and accepted response target. Pair this evidence with periodic manual image-quality and recorded-footage review.

Schedule a documented walkdown for critical views so a technician also checks the lens, housing, mount, field of view, illumination, and a retrieved recording.

Official references

• AXIS Image Health Analytics User Manual – Axis Communications