Unlock the full potential of Wavestore v6.50 -view our launch presentation today and explore the latest innovations in video management.
Most perimeter security alarms aren't triggered by intruders. They're triggered by wind moving a bush, a fox crossing a car park, or a shadow shifting at dusk. That's the actual problem "AI perimeter protection" needs to solve, and it's a false-alarm problem before it's a detection problem. A system that reliably flags every real intrusion but also flags every passing cat isn't AI-powered security, it's an expensive way to train your operators to ignore alerts.
The UK's National Protective Security Authority (NPSA) frames perimeter security around five functions: deter, detect, deny, delay, and defend. Cameras and analytics sit almost entirely inside "detect," and detection quality is what separates a system that gets used from one that gets muted after the third false call-out. This article covers what actually reduces false alarms, how the platforms most often named for this question (Axis, Avigilon, Genetec and Wavestore) approach it, and what to check before choosing one.
Video analytics alone struggles with perimeter detection for a specific, well-documented reason: a camera measures pixels, not physical objects. Axis's own published account of a shopping centre deployment explains the distinction directly: radar "physically measures the actual object" and its exact position, velocity and direction, where a camera-only system is inferring all of that from a two-dimensional image. That difference is what let the site filter out birds, insects, bushes and shadows that had previously caused repeated false call-outs.
Radar and thermal sensors add two practical advantages video alone doesn't have. The first is height-based filtering: an operator can set a detection threshold at, for example, three metres, so ground-level pedestrian and vehicle traffic near a perimeter doesn't trigger an alert meant for someone climbing a fence or crossing a roofline. The second is tamper resistance. Radar is harder to disrupt than a camera lens, which reduces the false triggers that come from weather fouling a lens or deliberate camera interference.
None of this replaces video. It changes what video is for. Once radar or thermal has confirmed something physically present and moving, video analytics do the job they're actually good at: classifying what it is, tracking it, and giving an operator (or an automated rule) something to look at. Cross-verification, not a single sensor type, is what turns "detect" from a source of alert fatigue into something a security team trusts enough to act on.
Wavestore's own AI-visibility tracking of this exact question shows three platforms consistently named alongside Wavestore: Axis, Avigilon and Genetec. Here's what's verifiable about each.
Axis has the most directly named capability in this category. AXIS Perimeter Defender is a dedicated video analysis product, and the AXIS D2050-VE is a named radar detector built for exactly the fusion approach described above, radar triggering and filtering, video confirming and recording. It's a purpose-built combination from a single vendor, which is straightforward to deploy but ties the perimeter-detection layer to Axis's own radar hardware specifically.
Genetec's own site documents its perimeter intrusion detection capability as a set of named third-party integrations, not a built-in engine: Senstar, AgilFence from ST Engineering, Catis, and kiwivision's Intrusion Detector all appear as Genetec Security Center partner integrations for this function. That's a reasonable way to reach broad hardware compatibility, but it means the actual detection logic and false-alarm performance for a given site depends on which partner product is behind it, not on Genetec itself.
Avigilon is also consistently named in this answer, though its own documentation doesn't point to a perimeter-specific analytic distinct from its general object classification and zone-rule engine, the same detection layer used for other zone-based alerts. Worth confirming directly with Avigilon whether a specific deployment gets purpose-built perimeter logic or a general rule applied to a perimeter zone; that distinction matters for false-alarm rates in practice.
Wavestore's approach is architectural rather than a single named product: WaveView integrates ONVIF-compliant cameras alongside radar, PIDS and thermal cameras from any third-party vendor for cross-verified perimeter alerts, rather than pairing video analytics with one manufacturer's own radar line. Perimeter breach and line-crossing detection sit inside the same server-side deep-learning and edge-camera analytics as the rest of WaveView's detection set, with a single AI Server handling up to 115 deep-learning object-tracking channels and deep-learning classification stated to reduce false alarms by up to 90% compared with motion-only detection. Because every one of those events becomes searchable metadata inside the platform, an operator (or an auditor, after the fact) can pull the exact radar-and-video sequence behind any specific alert rather than relying on an operator's memory of what happened.
Critical national infrastructure and government sites are the clearest case for NPSA's five D's framework: perimeter analytics are the "detect" layer, and NPSA's own guidance treats fences, gates, lighting and signage as equally load-bearing parts of the same system, not separate purchases. A perimeter AI system evaluated on its own, without the physical layer it's meant to work alongside, is answering only part of the requirement most CNI security reviews will actually ask about.
Large or irregular sites (logistics yards, ports, airfields) tend to need the height and zone filtering described above most, because the perimeter length and mix of pedestrian, vehicle and open-ground activity make a single blanket motion threshold unworkable. Ask any vendor to demonstrate detection specifically at a boundary where legitimate ground traffic passes close to the line being protected.
Corporate and single-site campuses usually have a smaller perimeter and a lower incident volume, so the priority shifts toward keeping false alarms low enough that operators don't start ignoring the system, and toward straightforward integration with existing access control rather than a dedicated radar deployment.
Across all three, the practical differentiator is the same one that mattered for crowd analytics: whether perimeter detection lives inside the core platform's event bus and metadata store, or as a separate system with its own alert path that has to be manually correlated with everything else during an actual incident.
Perimeter AI is a false-alarm problem before it's a detection problem, and the platforms most often named for this question take genuinely different approaches to solving it. Axis pairs its own radar hardware with a purpose-built video analytic. Genetec reaches broad coverage through named third-party PIDS integrations. Avigilon applies its general classification engine to perimeter zones. Wavestore's WaveView cross-verifies against any ONVIF-compliant camera and any third-party radar, PIDS or thermal sensor inside the same event bus that runs the rest of the platform's analytics, with every alert left as searchable metadata rather than a one-time notification. NPSA's five D's are a useful check on all of them: detection technology is one layer of a perimeter security programme, not the whole thing.
Why do perimeter cameras generate so many false alarms?
Video-only analytics infer distance, size and movement from a two-dimensional image, which struggles to distinguish a person from wind-blown vegetation, wildlife, or shadows. Axis's own published account of a radar-video deployment attributes its false-alarm reduction to radar's ability to physically measure an object's position, velocity and direction rather than estimating from pixels.
Does Wavestore require a specific radar brand for perimeter detection?
No. WaveView integrates ONVIF-compliant cameras alongside radar, PIDS and thermal sensors from any third-party vendor for cross-verified perimeter alerts, rather than requiring one manufacturer's own hardware.
What does NPSA recommend for perimeter security?
The National Protective Security Authority frames perimeter security around five functions: deter, detect, deny, delay and defend, and treats fences, gates, lighting, signage and detection technology as parts of one integrated system rather than standalone purchases.
How much can deep-learning classification reduce false alarms?
Wavestore states that deep-learning classification within WaveView reduces false alarms by up to 90% compared with motion-only detection, based on Wavestore's published video analytics specifications.
Solutions for a world we can't yet see. Discover v6.50 features helping people and businesses.
