Perimeter security architecture

Map the perimeter into detection zones and assessment areas. Define the purpose, response path and responsible operator for each sensor layer.

Overview#

A perimeter is a system of zones and response opportunities, not a fence line or a single sensor. Design each layer around a defined threat, operating context, detection objective, assessment path, response owner, and tolerated failure.

Layered model#

approach awareness -> boundary definition/deterrence -> detection
 -> localization/tracking -> independent assessment -> response decision
 -> delay/access control -> intervention -> recovery and learning

Record the boundary of responsibility: public space, controlled approach, outer perimeter, sterile zone, vehicle/pedestrian portals, building line, and protected asset. A sensor alarm near a mapped line doesn't prove crossing, intent, identity, or compromise.

Canonical zone and event model#

Use stable identities for site, perimeter segment, detection zone, sensor, camera view, illumination, portal, response sector, and maintenance exclusion. Geospatial coordinates and drawings need version and datum/reference where applicable.

Preserve source observations before correlation:

• source native event, object/track ID, location/zone, direction, time, confidence/quality, and classification;
• alarm, pre alarm, tamper, masking, fault, degraded, maintenance, bypass, restore, and unknown states;
• correlated incident ID, association rule/version, contributing and contradictory sources;
• operator assessment, response action, outcome, and closure reason.

Correlation reduces workload but can amplify common errors. Don't overwrite an original radar, fence, analytic, access, or camera event with one fused verdict.

Design dimensions#

Availability and health#

Monitor network reachability and the sensing performance that can be observed. Track enclosure/tamper, alignment, calibration age, obstruction/vegetation, signal/background/noise trends, power, battery, link/path, time, configuration drift, event flow, camera usability, recording, and operator destination.

Map common mode dependencies. Two detectors on one pole, power supply, switch, fibre route, wireless backhaul, analytics server, clock, or operations client don't provide independent assurance.

Maintenance needs controlled zones, approved test targets/stimuli, environmental coverage, alarm to assessment correlation, and proof that bypasses expire. Don't publish sensitive detection thresholds, blind spots, response times, or defeat details in general documentation.

Assessment and automation#

An associated video pane must show its camera identity, live/recorded status, image timestamp, clock uncertainty, recording gaps, view/occlusion limits, and analytic status. A cached or stale image must never appear as live.

Automatic camera cueing or track handoff may improve assessment but should bound movement, respect privacy masks/priority, surface loss of track, and not displace evidential recording. Automatic barrier or gate actuation from uncertain sensor/analytic output needs a separately approved safety and authorisation design.

Acceptance evidence#

• traceability from threat and zone objective to each sensor and assessment source;
• coverage/gap evidence across representative environmental and operating conditions;
• event location, timestamp, camera association, and response workflow;
• nuisance and missed detection review without unsupported universal performance claims;
• power/network/time/server/camera/sensor failure and recovery behaviour;
• bypass/maintenance controls and configuration/change audit;
• privacy boundaries, retention, export, and authorised disclosure.

Source baseline#

IEC 62676-4:2025 gives current application guidance for video surveillance systems. IEC 62676-6:2026 defines methods to evaluate and grade real time intelligent video analysis, including intrusion and complex scenarios under operational stress. Neither source establishes performance for an untested site or product.

See Perimeter and detection systems and Radar, fence, and buried detection.