Self-Learning Video Analytics

Share on facebook
Share on google
Share on twitter
Share on linkedin
computer screen showing analytical data for self learning video analytics

Security systems have come a long way. Now they teach themselves to recognise unusual events through pattern-based analytics and self-learning video tools. The ongoing learning process improves the system’s ability to recognise real from false alarms, makes security team surveillance more efficient in real time, and reduces costs involved in manually reviewing footage.

Self-learning video analytics essentially improves “event-based” surveillance by providing proactive real-time video monitoring, which enables your Gaming or Duty Manager to respond in real time to potential security risks. The VMS (Video Management System) basically uses AI to support a pattern-based learning system. This learn-by-example video analytics tool is designed to improve surveillance incident detection and if implemented correctly reduces the costs associated with time spent manually reviewing footage to find a particular event.

The advanced pattern-based analytics of self-learning video tools can accurately recognise people and vehicle movements, whilst ignoring any activity not relevant to the scene. By using one’s own embedded cameras the system is constantly learning and storing information. This ongoing process ensures that there is a reduction in false positives and that security alerts are genuine in nature, thus reducing time wasted on investigating non-issues.

These systems use learn-by-example object classification technology to provide users with the capability to provide feedback into the system regarding the accuracy of alerted events. The system in turn, is able to learn where it went wrong and make more accurate assessments in future.  This feedback loop is constantly training the system and thus improves the system’s ability to recognise real from false alarms. Over time, this further improves the efficiency and accuracy of your surveillance operations through the prioritisation of events and alarms based on user inputs and feedback.

These systems are smart, cost effective and easy to use. Their intuitive User Interface combined with robust video analytics capabilities make them an asset for any Club or Venue. Real-time events and forensic analysis ensure that the system identifies and notifies you of scene changes, missing objects and any rule violations. These systems can be implemented across many devices, while providing precise control of event playback which allows for incredibly efficient footage retrieval.

Most common examples:

  • OBJECTS (e.g. people or vehicles) IN AREA The event is triggered when the selected number of objects are present in the region of interest. The object can appear from within the region of interest or enter from outside the region of interest.
  • OBJECT LOITERING The event is triggered for each object that stays within the region of interest for an extended amount of time.
  • OBJECTS CROSSING BEAM The event is triggered when the specified number of objects have crossed the directional beam that is configured over the camera’s field of view in the selected time period. The beam can be unidirectional or bidirectional.
  • OBJECT APPEARS OR ENTERS AREA The event is triggered by each object present in the region of interest. The object can appear from within the region of interest or enter from outside the region of interest.
  • OBJECT NOT PRESENT IN AREA The event is triggered when no objects are present in the region of interest.
  • OBJECTS ENTER AREA The event is triggered when the specified number of objects have entered the region of interest from outside of the region.
  • OBJECTS LEAVE AREA The event is triggered when the specified number of objects have left the region of interest.
  • OBJECT STOPS IN AREA The event is triggered for each object in a region of interest that stops moving for the specified threshold time.
  • DIRECTION VIOLATED The event is triggered for each object that moves in the prohibited direction of travel.
  • CAMERA TAMPERING The event is triggered when the scene unexpectedly changes.

If they can teach a car to drive itself using pattern-based analytics and teach-by-example technology,  then a security system that learns to recognise unusual events in real-time, continually improves the system’s ability to recognise real from false alarms, and reduces costs involved in manually reviewing footage to find a particular event, is the most practical application of complicated technology you will ever find!

To find out more contact Wes Stevens

Sales Manager – Element Security

t: 1300 325 276

e: wes@elementsecurity.com.au