Genetec Security Center AI Integration Guide
Why a Genetec-specific guide
Genetec Security Center is a unified platform that many transit agencies and large industrial sites already run for video, access control, and ALPR. The integration question its operators ask is specific: how do we add real safety analytics without forklifting the cameras, retraining the control room, or moving video off premise? This guide answers that for Security Center the same way our Milestone checklist does for XProtect. DHI sits beside Security Center as an edge-native analytics layer, not as a replacement for it.
The integration goal is simple. Operators keep living in Security Center. DHI does the inference at the edge and hands Security Center clean, structured safety events it already knows how to route.
Stage 1: Edge Node and Stream Access
Stand up a DHI edge node on the same segment as your Security Center archiver and grant it read access to the camera RTSP substreams. DHI does not need the recording path or any control-plane access. It only needs to see the video, classify locally, and emit events. Because inference happens on the node, video never leaves the building and detection stays inside 150 milliseconds end to end.
Confirm ONVIF profile support
Most Security Center camera integrations are ONVIF Profile S or Profile T. Confirm the cameras you are targeting expose ONVIF events, because that is the channel DHI uses to deliver incident metadata that Security Center can natively ingest.
Plan the safety VLAN
As with any vision deployment, isolate the analytics traffic on a dedicated VLAN. This keeps high-bitrate inference traffic away from access-control and ALPR paths that share the platform.
Stage 2: Event-to-Alarm Configuration
In Security Center, model each DHI incident class as an event source, then attach it to an event-to-action rule. The action you almost always want is raise an alarm, switch the operator monitor to the relevant camera, and bookmark the archive so the clip is one click away during review.
Map severity deliberately
Security Center alarm priorities should mirror real operational urgency. A confirmed track-trespass or person-down event belongs at the top of the dispatcher queue. A lower-confidence loitering signal can sit lower. Mapping DHI confidence to Security Center priority is what turns a feed of detections into a triaged worklist.
Preserve the audit trail
Because every DHI event carries a timestamp, a camera reference, and a confidence value, the Security Center archive becomes a defensible record of what was detected and when. That matters for incident review and for compliance reporting after the fact.
Stage 3: Validate Against Real Scenes
Trigger each incident class in the field and confirm three things in the control room: the alarm fires in Security Center, the right camera surfaces to the operator, and the archived clip is bookmarked at the moment of detection. Run this for every camera in the pilot scope, not just one, because angle and lighting change the picture camera to camera.
Common first-week findings
The two issues that surface most often are a confidence threshold set too low for a busy scene, and a camera mounted at an angle that clips the detection zone. Both are quick to correct and neither is an integration fault. Once corrected, the expected outcome is the same one our other deployments show: dispatchers receive far fewer, far more actionable alerts than native motion detection ever produced.