When Everything Triggers an Alert, Nothing Does
Ask a security manager about their perimeter system and you rarely hear about the intrusions. You hear about the nights it went off eleven times for nothing.
The Real Cost Isn't the Noise — It's the Alert That Gets Ignored
A team that has dismissed forty nuisance alerts in a week responds slower to the forty-first, and eventually stops looking. Then someone disarms the zone that keeps firing. The system stops being a tool and becomes background noise.
Basic Motion Detection Cannot Tell What Moved
It alerts on pixel change — which is why a moth on the lens, a shadow crossing at dawn, or a plastic sheet flapping in the wind all look identical to a person climbing a fence.
Indian Sites Have Their Own List — And Most Vendors Never Name It
Stray dogs and cattle wander plant perimeters and warehouse yards routinely. Monsoon rain and wind move vegetation across camera views for weeks. Dust hangs in the air at industrial sites, winter fog blankets northern ones, and insects swarm infrared illuminators after dark. Every one looks like motion to a system that only sees motion.
Long Perimeters, Empty Sites, and Hardware Projects
A guard cannot watch every metre of fence line continuously. Solar farms, substations, yards and construction sites sit empty overnight — precisely when they're at risk. And fence sensors, buried cable and beam systems each mean survey, trenching or fence works and per-metre cost before anyone sees a single image.
What Is AI Video Intrusion Detection?
Classification, not pixel change. Motion detection asks did anything change? Object detection asks what is it? — identifying a person, a vehicle or an animal, and only then applying your rules. That difference is what separates a system your team trusts from one they mute.
Lines, zones and rules
A virtual line (or tripwire) is drawn along a boundary — crossing it raises an event, and rules can require a direction so people walking past the fence don't trigger what only inward movement should. A zone is an area where presence itself is the event. Rules decide what actually alerts: which object classes, which direction, which zone, at what times.
Honest about the limits
Classification substantially reduces nuisance alarms but does not eliminate false positives, and detection depends on what the camera can actually see — distance, angle, lighting and weather all affect it. We publish no accuracy percentage, because you have probably been shown one before. The comparison below is specific about where other technologies do this better.
Where it fits
It runs on the Tentovision AI Video Analytics platform and its video management software, hosted on AWS with on-site options, alongside number-plate recognition on the same feeds.
Physical intrusion detection — not network IDS
Worth clearing up immediately, because the shared name causes genuine confusion. In cybersecurity, an "intrusion detection system" or IDS monitors network traffic for malicious activity — a different field, a different team, a different problem. This page is about physical intrusion detection: software watching camera feeds for people or vehicles entering areas they shouldn't.
Where it sits in layered security: the security profession describes perimeter protection through the five Ds — deter, detect, deny, delay and defend. This is squarely a detect layer. It doesn't stop anyone; it tells you, early and with a picture, that someone is there.
What the System Detects, Decides and Does
Eight capabilities, running on the cameras already covering your site. (Every capability is configured per site; availability is confirmed at assessment.)
a site boundary
Virtual Lines & Tripwires
Lines drawn on the camera view along boundaries, gates or approach routes, with rules that can be configured for direction — so an inward crossing raises an event while movement along or away from the line does not. The difference between alerting on a perimeter breach and alerting on the public footpath outside it.
a compound
Restricted-Zone Detection
Areas drawn on the camera view where presence itself is the event — a substation compound, a chemical store, a rooftop, a yard after close. Multiple zones can be defined per camera, each with its own rules, so a busy loading area and a permanently off-limits enclosure aren't treated the same way.
classified and filtered
Object Classification — Person, Vehicle, Animal
The capability that makes the rest usable. The system identifies what triggered a detection and alerts only on the classes you care about — typically people and vehicles — while stray dogs, cattle and moving foliage are recognised and set aside. This is what reduces nuisance alarms to a level a team will actually keep responding to.
after working hours
Time-Based Arming Rules
Zones can be configured to arm and disarm on a schedule, so an area that's legitimately busy during shifts becomes a no-entry zone after close. Shift-aware rules mean the system reflects how the site actually operates rather than treating 3 p.m. and 3 a.m. identically.
of what triggered it
Alerts With Visual Verification
Every alert can carry a snapshot or clip showing what triggered it, delivered to the channels your team already uses. This matters more than it sounds: an alert with an image can be assessed in two seconds, while one without requires someone to go and look — which is how response times slip.
triggered at the boundary
Deterrent Output — Siren, Strobe, Voice-Down
Detections can typically be configured to trigger connected devices: a siren, a strobe, or a speaker for a live or pre-recorded announcement. Challenging someone at the perimeter is often what turns an intrusion into an aborted attempt, rather than an incident you review afterwards.
response per zone
Multi-Zone Rules With Per-Zone Response
Different areas warrant different responses. A perimeter line might raise a control-room alert; a critical enclosure might additionally trigger a siren and an escalation. Rules and outputs can be configured per zone rather than applied uniformly across the site.
by zone and time
Evidence Capture & Incident Log
Each event can be logged with time, location and footage, building a searchable record for investigation, insurance and internal review — and, over time, showing which zones and hours actually generate activity.
On limits, plainly: classification substantially reduces nuisance alarms, but no system eliminates false positives entirely, and any vendor promising that is overselling. Detection also depends on what the camera can actually see — distance, angle, lighting and weather all affect it. The comparison below is specific about the conditions where other technologies do this better.
Vendor-Agnostic · 200+ Camera Brands Supported
Tentovision connects to any IP or CCTV camera over ONVIF, RTSP, and standard NVR APIs, streaming via the Cloud Adapter where needed — no trenching, no fence works, no per-metre sensor cost. (What matters is whether cameras actually cover the lines and zones you want protected, at a workable distance and angle — confirmed by a perimeter assessment.)
How It Works
Five steps, on cameras already covering your site. (Rules, outputs and integrations are configured per site and confirmed at assessment.)
Define Lines & Zones
Virtual lines and zones are drawn directly on each camera's view — along the fence line, across gates, around restricted enclosures — using cameras already covering those areas.
Detect & Classify
The AI analyses the feed and identifies objects entering the defined areas, classifying them by type rather than simply registering that something moved.
Evaluate the Rules
Each detection is checked against that zone's rules: right object class? Right direction? Inside armed hours? Most nuisance events are filtered here, before anyone is interrupted.
Alert & Respond
Qualifying events raise an alert with visual verification to your team's channels, and can trigger connected outputs — siren, strobe or voice-down — so the site responds immediately rather than only recording.
Log the Incident
Events are recorded with time, zone and footage in the command centre on AWS cloud or on-site, with camera health monitoring flagging any camera that goes offline.
AI Video vs Motion Detection vs Fence Sensors vs Radar vs Thermal
Perimeter detection is a genuine technology choice, and the right answer depends on your site — fence structure, perimeter length, visibility conditions, and how much nuisance alarm your team will tolerate. Here is an honest view, including where other approaches beat this one. (Characteristics are typical; verify against quotes for your site.)
| Dimension | AI video (Tentovision)Existing cameras | Basic motionPixel change | Fence sensorsContact | RadarRF sensing | ThermalHeat imaging |
|---|---|---|---|---|---|
| Identifies what triggered it | ✓ person / vehicle / animal | ✗ motion only | ✗ contact only | ✗ can't identify | ~ heat signature |
| Visual verification | ✓ image with every alert | ~ | ✗ | ✗ | ✓ |
| Nuisance-alarm profile | ✓ low with classification | ✗ very high | ~ high from wind, vegetation | ✓ low | ✓ low |
| Detects cutting / climbing at the fence | ✗ no | ✗ | ✓ their strength | ✗ | ✗ |
| Fog, heavy rain, dust | ✗ degrades | ✗ | ✓ unaffected | ✓ strong | ✓ strong |
| Full darkness | ~ needs IR or lighting | ~ | ✓ | ✓ | ✓ their strength |
| Large open areas, long range | ~ camera-dependent | ✗ | ✗ fence line only | ✓ wide coverage | ✓ |
| Uses existing cameras | ✓ no new hardware | ✓ | ✗ install per metre | ✗ dedicated units | ✗ dedicated cameras |
| Installation | ✓ software | ✓ | ✗ trenching / fence works | ~ mounting + calibration | ~ camera replacement |
| Best fit | Classified detection with visual proof on cameras you already have | Low-risk areas | Fence-line contact detection | Large open sites, poor visibility | Darkness & adverse weather |
The honest read. If your site faces persistent fog, dust or monsoon conditions and you need guaranteed detection through them, thermal or radar is the right investment, and we'd tell you so. If your threat is someone cutting or climbing the fence itself, fence sensors detect physical contact that no camera can see. Where AI video is strongest is knowing what crossed the line and being able to see it — classification plus visual verification, on cameras that are already installed. On serious perimeters these layers are often combined, with video providing the verification that radar and fence sensors cannot.
Where It's Used
Perimeter & Fence Lines
Virtual lines along the boundary of a plant, yard or compound, with directional rules so passers-by outside the fence don't generate alerts.
Restricted Zones Inside the Site
Substation compounds, chemical and fuel stores, transformer yards, server rooms — anywhere presence alone is the event, regardless of how someone got there.
After-Hours Premises
Warehouses, workshops, offices and retail back-of-house, where the whole area becomes a no-entry zone once the shift ends and the arming schedule takes over.
Gates & Entry Points
Approach routes and vehicle gates, where detection pairs with number-plate recognition to answer both someone is here and this is which vehicle.
Remote & Unmanned Sites
Solar farms, substations, telecom sites, storage yards and construction sites — empty overnight, with nobody present to notice anything until morning.
Rooftops, Service Areas & Blind Spots
The parts of a site that are rarely walked and rarely watched, where an intrusion can go unobserved for hours.
Industries We Serve
Manufacturing Plants
Perimeter lines, gates and restricted process areas across large sites where the fence runs for kilometres and staffing is concentrated elsewhere — alongside other analytics on the same cameras across manufacturing operations.
Warehousing & Logistics
Yards, loading areas and stock compounds, where high-value goods sit outside working hours across warehouse sites.
Solar Farms, Substations & Energy
Large, remote, unmanned sites with valuable equipment and long perimeters — one of the fastest-growing applications for perimeter detection globally.
Construction Sites
Temporary perimeters, plant and materials on site overnight, and a boundary that changes as the project progresses.
Data Centres & Critical Infrastructure
Layered perimeters where detection feeds a defined escalation process and every event needs an evidence trail.
Campuses & Commercial Premises
Educational, institutional and commercial sites with open boundaries during the day and restricted access after hours.
Why Tentovision for Intrusion Detection
Runs on the Cameras You Already Have
No trenching, no fence works, no per-metre sensor cost, no camera replacement — the detection layer goes on existing CCTV, which is usually what makes the project approvable at all.
Classification Is the Point, Not a Feature
The reason to move from motion detection to AI detection is that your team keeps responding to alerts. Everything else follows from that.
A Failed Camera Doesn't Become a Silent Blind Spot
Camera health monitoring flags offline or degraded cameras — on a perimeter, an unnoticed dead camera is a gap nobody knows about.
One Platform, One Set of Cameras
Intrusion detection sits on the AI video analytics platform and its video management software, alongside number-plate recognition and other analytics on the same feeds.
Honest About Fit
We publish where thermal, radar and fence sensors do this better — and we don't quote a false-alarm percentage, because you've probably been shown one before.
Or Let Us Watch It Instead
If you'd rather not run it yourself, our e-surveillance service is the alternative: our operators watch your cameras, verify alerts and respond on your behalf. This page is the software your own team runs; that is the managed service. Same detection underneath — the difference is who is watching.
India-based engineering and support, with detection validated on your own cameras and conditions before rollout.
Intrusion Detection in the Tentovision Security Suite
Detection is one layer. Identifying the vehicle, keeping cameras alive, and having someone watch are others — all on the same feeds.
Video Intrusion Detection You Are Here
The detection software your own team runs · virtual lines & zones · person/vehicle/animal classification · visual verification
E-Surveillance Service Managed Alternative
The same detection, but our operators watch your cameras, verify alerts and respond on your behalf. The difference is who is watching.
AI Video Analytics Parent Platform
The umbrella — 20+ analytics modules on the same infrastructure
ANPR / License Plate Recognition
Answers which vehicle at gates, where intrusion detection answers someone is here
Camera Health Monitoring
Flags offline or degraded cameras — an unnoticed dead camera is a silent gap in the perimeter
CCTV Command Center
Multi-site view where perimeter events across every location are monitored together
Every Tentovision module runs on the same camera feeds and the same dashboard. Add a module in software — no new hardware.
What Security Teams Ask First
Answers written to be useful for security and facility managers — accurate enough to be cited by ChatGPT, Gemini, and Perplexity when they research this topic.
What is video intrusion detection?
Video intrusion detection is AI software that watches a camera's view for unauthorised entry into a defined area — a virtual line across a boundary, a restricted zone, or a whole site outside working hours — and alerts the moment something crosses it. Unlike basic motion detection, it can classify what triggered the alert, so a person climbing a fence is treated differently from a dog crossing the yard.
Is this the same as an intrusion detection system (IDS) in cybersecurity?
No, and the shared name causes real confusion. In cybersecurity, an intrusion detection system monitors network traffic for malicious activity. This is physical intrusion detection: software watching camera feeds for people or vehicles entering areas they shouldn't. The two share a name and nothing else — different technology, different teams, different problems.
How do you reduce false alarms?
Primarily through object classification: instead of alerting on any movement, the system identifies what moved and alerts only on the classes you care about, typically people and vehicles, while animals and vegetation are set aside. Directional rules, minimum object size, zone definitions and time-based arming reduce noise further. No system eliminates false positives entirely — anything that promises zero is overselling — but classification is the difference between alerts your team acts on and alerts they mute.
Can it tell the difference between a person and an animal?
Yes — that's the core of how nuisance alarms are controlled. On Indian sites it matters more than most vendors acknowledge: stray dogs and cattle are among the most common causes of false alerts on plant and warehouse perimeters. Accuracy depends on camera view, distance and lighting, and is confirmed on your own site rather than promised in advance.
Does it work at night, in fog, or in heavy rain?
At night it works where cameras have infrared or adequate lighting, which most perimeter cameras do. In fog, heavy rain and dust, visible-light video degrades — and thermal cameras and radar genuinely perform better in those conditions. If your site faces persistent fog or monsoon conditions and needs guaranteed detection through them, a thermal or radar layer is the right investment, with video AI as a complement rather than a replacement.
Can I use my existing CCTV cameras?
In most cases yes, and that's the main economic argument. The software runs on the IP/CCTV cameras already covering your perimeter and entry points, so there's no trenching, no fence works and no per-metre sensor cost. What matters is whether cameras actually cover the lines and zones you want protected, at a workable distance and angle — a perimeter assessment identifies which existing views work and where coverage is genuinely missing.
How is this different from the motion detection already on my cameras?
Motion detection alerts on pixel change, which is why it fires for wind, shadows, rain, insects and animals. AI intrusion detection identifies what moved before deciding whether to alert, and applies rules about direction, zone, object class and time of day. The practical difference is the number of alerts your team has to look at each night.
How does it compare to fence sensors, radar or thermal?
Each detects something different. Fence sensors detect physical contact — cutting, climbing, impact — which video cannot, though they carry a high nuisance rate from wind and vegetation. Radar covers large open areas and works in poor visibility, but cannot identify what it detected. Thermal sees in darkness and through some weather. Video AI's advantage is classification and visual verification. Serious perimeters often layer two or more.
Can I set different rules for different areas and times?
Yes — rules can be configured per zone and per schedule. A perimeter line might alert only on inward crossings by people or vehicles; a critical enclosure might alert on any presence at any hour; a loading yard might be unmonitored during shifts and armed after close. Which rule types are available for your deployment is confirmed at assessment.
Can it trigger a siren, strobe or voice announcement?
It can typically be configured to trigger connected outputs, so an intruder is challenged at the perimeter rather than simply recorded. Being addressed at the boundary is often what ends an attempt before it becomes an incident. What's possible depends on the devices installed on your site and is scoped during assessment.
How is this different from your e-surveillance service?
This is the detection software, which your own team operates and responds to: alerts go to your security staff, your control room or your phones. Our e-surveillance service is different — our operators watch your cameras, verify alerts and act on them on your behalf. Same detection technology underneath; the difference is who is watching and who responds. Some sites run the software in-house; others prefer the service.
Which is the best intrusion detection system?
Any honest answer starts with the site: perimeter length, whether there's a fence to instrument, visibility conditions, how much nuisance alarm the team will tolerate, and whether alerts will be visually verified. Fence sensors, radar, thermal and video AI each suit different sites, and large perimeters often combine them. Tentovision is the video-AI layer — classification-based detection on cameras you already own, with visual verification. The practical way to decide is a perimeter assessment.