Skip to content
NewNow in GermanAuf Deutsch
Camzify
AI detection · Line intrusion detection

Line intrusion detection

Line intrusion detection is a virtual tripwire placed across any area in the camera view, with directional control. It fires when a confirmed object track crosses the line — not when a shadow moves or a light flickers. This is the foundation of perimeter security in Camzify.

  • People crossing a perimeter fence line
  • Vehicles entering a restricted loading zone
  • Directional crossings at one-way gates or turnstiles
The console live view with a line intrusion alert drawn on the camera frame
Line intrusion detection

This capability detects and alerts on:

  • • People crossing a perimeter fence line
  • • Vehicles entering a restricted loading zone
  • • Directional crossings at one-way gates or turnstiles
  • • After-hours movement across secured boundaries
  • • Repeated crossings at the same line within a short window

Why line intrusion detection matters

A perimeter fence or boundary line only works as a deterrent if someone, or something, is watching it continuously. A guard on a walking patrol covers a boundary for a few minutes every hour at best; the rest of the time, the line is unmonitored.

Fixed motion sensors and basic pixel-change alarms fill some of that gap, but they trigger on anything that moves in frame, a delivery truck passing on the street outside a fence, a tree branch in the wind, a cat crossing the yard. Security teams either drown in false alerts or tune sensitivity down until real crossings get missed too.

Line intrusion detection solves both problems at once: it watches the boundary every second of every day, and it only fires when a confirmed person or vehicle track actually crosses the line in the direction that matters.

A sequence of frames from one camera showing a line intrusion event build up to the alert
Tripwire logic

How it works

Detecting the crossing

The model uses multi-object tracking to maintain a persistent identity for every subject in the frame, frame over frame. A single confirmed track — not a single frame of pixel change — is what the line rule evaluates.

Directional logic

When a tracked object's trajectory intersects the defined line, the system checks the direction of travel against the configured rule. A line set to alert only on inbound crossings ignores a subject walking away from the boundary, cutting the alert volume roughly in half at most perimeter points.

Alert delivery

Each alert includes the object type, confidence score, and timestamp.

  • When AI attribute extraction is enabled, the alert also carries structured attributes like clothing color and behavior description.
  • Alerts route through the platform's notification system and appear in the notification queue.
  • Each queued alert shows its severity, its acknowledgment status, and the option to mark it as a false positive.

Configuration

Lines are drawn directly on the camera view in the configuration panel. A line across an entrance also counts people entering and leaving separately, the basis of counting foot traffic on existing cameras. Each line supports:

  • • Directional control. A→B, B→A, or bidirectional
  • • Sensitivity adjustment per line
  • • Notification window per camera, e.g. notify after hours only
  • • Object-class filtering, e.g. people and vehicles only
  • • Multiple lines per camera, each with independent rules
Configuration panel showing a directional line drawn across a camera view with sensitivity and schedule controls
Line configuration
Camera scenes across several sites where line intrusion detection is used
Perimeter monitoring

Common scenarios

  • • A perimeter fence line monitored after business hours, alerting only on inbound crossings
  • • A loading dock lane where vehicle crossings outside scheduled delivery windows trigger an alert
  • • A rooftop access hatch where any crossing at all is treated as non-compliant
  • • A parking structure boundary that separates public and staff-only areas
  • • A construction site fence line active only overnight and on weekends

In a patrol round

During a virtual patrol round, line intrusion alerts contribute to the compliance assessment at each camera stop. If an alert fired between rounds, it is logged alongside the checklist results in the patrol report.

The footage around each alert is kept under the camera's own retention window on cloud video surveillance, so the clip and the recording it came from are on one account.

FAQ

Frequently asked questions

Line intrusion detection, also called line crossing detection or virtual tripwire, is an AI capability that triggers an alert when a tracked object crosses a defined line in the camera view. Unlike pixel-based motion detection, it operates on confirmed object tracks, meaning shadows, lighting changes, and camera noise do not trigger false alarms.

Yes. Each line can be configured for directional control, left-to-right, right-to-left, or both directions. This is essential for scenarios like one-way gates, loading dock entrances, or perimeter fences where you only care about inbound crossings.

Motion detection responds to pixel changes in the frame. Line intrusion detection responds to a confirmed object track crossing a specific line. The difference is precision, motion detection catches everything including irrelevant movement; line intrusion catches only what matters.

A single camera can support multiple lines, each with its own direction, sensitivity, and schedule. A wide-angle view of a fence line, for example, can carry separate lines for a vehicle gate and a pedestrian gap without needing a second camera.

The underlying object tracker is trained to distinguish people, vehicles, and other confirmed object classes from environmental movement like blowing leaves, rain, or small animals. Object classes to alert on can also be restricted per line, so a perimeter line can be set to ignore anything that isn't a person or vehicle.

Detection quality at night depends on the camera's own low-light or infrared performance rather than the detection model itself. Any camera feed with enough visibility for a human reviewer to identify a person or vehicle is enough for the tracker to build a confirmed object track.

See the detections on live cameras

A 15-minute demo of the live product: the detections your site needs, running on real camera feeds, and a quote per camera for your sites.