Home / Can your factory CCTV support video analytics? A feasibility checklist
Can your factory CCTV support video analytics? A feasibility checklist
Start with one visible event and a reliable way to judge whether it occurred. Buying more cameras or choosing a model comes later. A useful feasibility review asks what the existing view can support, under the conditions in which the factory actually operates.
Describe an observation, not a broad ambition
“Improve productivity” is a business objective. “Identify when a defined machine appears inactive during a scheduled operating period” is a candidate observation. Write down the event boundary, exclusions and intended response. Planned breaks, changeovers, cleaning and material shortages should not silently become the same category of idle time.
Inspect visibility across operating conditions
Review actual footage from the relevant shifts. Check whether the important action is large enough in the frame, whether people or material block it, and whether lighting changes alter what can be seen. Include camera movement, missing footage and network interruptions in the discussion. A clear demonstration clip is insufficient evidence of routine reliability.
Some outcomes are not visible. A camera cannot establish hidden material defects, electrical consumption or the reason a worker paused simply because it sees an object or person. Decide which observations require equipment records, meter readings or supervisor input.
Build a representative evaluation set
Choose examples that include normal activity, the target event and confusing near-misses. Keep a separate set for evaluation rather than tuning on every example used to report results. Record who reviewed the reference labels and how disagreements were resolved. In time-based use cases, define acceptable timestamp tolerance and minimum event duration before scoring.
Measure misses and false alerts separately
A pilot report should explain how many real events were found, how many were missed and how many alerts were incorrect. A single headline accuracy percentage can hide important trade-offs. Also assess whether evidence is available when the supervisor needs it and whether the alert leads to a useful action.
Assign a person and a process
Agree who reviews an alert, how they correct it, and what happens when evidence is missing. Define access, retention and the permitted use of footage before collecting pilot material. Evaluate privacy and workplace implications with the responsible team. These are design inputs, not paperwork to add after the system is built.
Make the expansion decision explicit
At the end of the pilot, review performance by camera and condition. Expand only to views and workflows for which the evidence supports the agreed criteria. A successful pilot at one workstation does not establish performance across a whole factory.
Viraya's worksheet provides fields for the observation, camera limitations, evidence set and review responsibilities. It is a planning aid; it does not claim that a camera or use case has passed an assessment.
Use the CCTV feasibility worksheet
Sources and scope
This article explains a review approach; it does not report a customer deployment or guaranteed result.