Specify what the camera must show
Describe an observable event, a location and the person responsible for reviewing it. “Monitor safety” is too broad to assess. A more useful starting point is a visible helmet requirement at a defined entrance during specified shifts, or a vehicle crossing an agreed gate boundary.
Check whether the view shows the relevant detail throughout the event. A wide yard camera may be appropriate for general movement but unsuitable for reading a plate. A checkpoint intended for PPE review needs a view of the relevant item, not merely a person entering the scene.
Review normal and difficult capture conditions
Collect authorized samples from the operating conditions the system will encounter. Include day and night lighting, glare, motion, occlusion, weather where relevant and changes between shifts. Record the camera position and scene conditions with each sample.
Review field of view, subject size, mounting stability and lighting before assuming a different model will solve a capture problem. Document which views are suitable, which require adjustment and which events cannot be reliably assessed from the available footage.
Check the actual stream and device configuration
Record camera and recorder models, firmware, available stream interfaces, codec settings and authorized access arrangements. Test a stream from the proposed processing location using the configuration intended for deployment.
ONVIF Profile T describes features for IP video interoperability, including H.264/H.265 streaming and metadata. Profile conformance should be checked for the actual device and client; it does not establish that a scene is suitable for a particular analytics task. Keep protocol compatibility and detection suitability as separate assessment results.
Size the complete data path
Specify the selected analytics channels, their capture settings and operating schedules. Installed camera count and active analytics channel count are different scope items. Measure the proposed configuration rather than extrapolating from an unrelated demonstration.
For on-site, private-cloud or hybrid processing, identify where video, snapshots and event records travel and where they are retained. Check network capacity, compute resources, storage where required, access restrictions and recovery behavior. Include camera loss and a processing restart in the readiness review.
Agree the event review workflow
An alert needs a destination, an owner and enough evidence for a decision. Agree the camera name, location, timestamp and event reference that operators will receive, along with the acknowledgement and escalation process.
Produce a readiness record listing reusable equipment, required capture changes, interface findings, infrastructure assumptions and a proposed pilot. Distinguish verified results from items still awaiting access or representative footage. This gives implementation and operations teams a common starting point.
Implementation checklist
- Select an observable event and its reviewer.
- Record camera, recorder and stream details.
- Assess day/night views and difficult conditions.
- Separate streaming compatibility from detection suitability.
- Map processing, evidence storage and network requirements.
- List capture changes and unresolved assumptions.
Technical references
Related technical guides
- Connecting AI agents to enterprise systems
- How to evaluate an industrial video analytics pilot
- An Odoo data migration checklist for implementation teams