Object-based Zone
This security zone type is available on the QbProtect product. |
The object-based security zone uses detected objects and specific object attributes (e.g., size, lifetime, traveled distance) within the zone’s volume to generate an alarm.
Application
- Advanced alarm generation
-
The object detection and tracking involved in generating the alarms within this zone enable a large number of possibilities. Proper parameterization can be used to configure, filter, and tailor the alarm generation in order to reduce the number of false alarms.
Parameters
See configuration API definition.
Intrusion conditions
- Alarm sizes (default: LARGE, MEDIUM)
-
Filter objects based on size. Only objects of a selected size category can raise an alarm. The selection of multiple object size categories is allowed.
-
LARGE: Cars, groups of people or large animals (e.g., deer).
-
MEDIUM: Non-occluded individual humans, standing or crawling.
-
SMALL: Crouching humans, partially occluded persons or small animals (e.g., rabbits).
The size category boundaries are configurable and can be adjusted.
-
- Minimum intruder lifetime (default: 0, unit: \(s\))
-
The minimum lifetime of the detected object before it can trigger an alarm.
- Minimum intruder track length (default: 0, unit: \(m\))
-
The minimum length of the object’s track before it can trigger an alarm.
- Objects in background (default: false)
-
Trigger alarms also for objects with the property
in_background
set. Enabling this option might increase the false alarm rate. - Minimum intruding duration (default: 0, unit: \(s\))
-
The duration for which all intrusion conditions have to be met until an alarm is raised. A value of 0 (the default) raises the alarm immediately upon first detection.
- Intrusion confidence (default: 100, unit: \(\%\))
-
The confidence threshold for raising an alarm. The current confidence value is computed over the time interval specified as the minimum intruding duration. A value of \(100\%\) ensures, that all intrusion conditions are met in all observations within the minimum intruding duration. Values lower than \(100\%\) make alarm generation more sensitive by allowing the configured ratio of observations within the minimum intruding duration window where all intrusion conditions are met.
The intrusion confidence is only configurable when the minimum intruding duration is nonzero.
Algorithm
The algorithm uses the tracked objects from the object tracking as input data. It operates in two stages:
Filter intruding objects
This step finds the subset of all currently tracked objects which are currently intruding into the zone.
An object is considered to be intruding when the object’s centroid is located within the zone’s cuboid. In order to narrow down the number of relevant objects the parameters for alarm sizes, minimum intruder lifetime, the minimum intruder track length and the in-background property are applied.
The alarm inside the intrusion state is set to \(true\) whenever the list of intruding objects is not empty.
Alarm filtering
Based on the intrusion state, the actual alarm state is determined. In this step, the configuration values for minimum intruding duration and intrusion confidence are applied.
Output data
See full API definition.
State
- alarm (\(true,false\))
-
The current alarm. A \(true\) value indicates an alarm.
- intrusion
-
Unfiltered pre-alarm state from the zone’s internal alarm logic.
- objects
-
List of objects currently intruding the zone. Each entry also contains a discretized representation of the track history and a snapshot of the current point cloud.
- intrusion confidence
-
Current confidence based on the minimum intruding duration window.