Wavestore’s Deep Learning Video Analytics adds the ability to detect and search for a wide range of behaviours and events, system-wide, to provide a truly intelligent Video Management System. This powerful extension to our core VMS software utilises virtually any IP camera and can easily be deployed as part of a new project, or retro-fitted to existing systems.
Wavestore Video Analytics have been designed to reliably detect humans and vehicles as well as alert operators to exceptional events. This is achieved through pre-configured rules that are set up using the on-board rules creation engine, via a simple-to-use web interface. Logical rules can be configured for single or multiple sets of analytics which, when triggered, perform a wide variety of actions via Wavestore’s dynamic event rules capability.
Up to 40 zones and tripwires per channel can be set to detect any activity either in an area or crossing a point of interest.
A robust object tracking engine tracks through occlusions to give an exceptionally high detection rate with very low false positives. Multiple modes permit optimization for intrusion detection or business intelligence tracking.
Count Lines quantify the number of objects in a zone or crossing a line, along with the option to show this count on-screen. This includes a unique calibration tool to separate individual targets and generate accurate counts.
A speed filter detects and displays the speed of a person or vehicle in a zone or crossing a line.
There is a comprehensive set of rules to detect object appearance or disappearance, entry and exit, stopping or dwelling, direction, tailgating and, of course, removed or abandoned object.
Because humans and vehicles are usually the only things of interest, deep learning has been deployed within Wavestore Video Analytics to filter out environmental and other unwanted activity. As such, activity from things like moving trees, clouds, car headlight reflections and animals etc., which can all cause false positive readings, are identified through deep learning, and ignored by the analytic algorithms.
Wavestore Video Analytics are deployed using our own Analytics Servers which are available in two versions, able to accommodate up to 32 and 64 analytics channels. They can run up to four different types of analytics simultaneously on each channel.
Tamper detection, designed to alert operators of unauthorised camera obstruction or movement, and image stabilisation which is used to improve image quality, are included as standard.
To improve detection accuracy across the entire scene, as well as minimise false alarms and speed up deployment, a built-in 3D calibration tool is used on set-up. This is to define the height, angle, and perspective of subjects over various distances within the field of view. The metadata from the Video Analytics Appliance can also be collected, via an API, for reporting purposes.
Video channels are restreamed to Wavestore Video Management Software (VMS) storage/ recording servers together with metadata which can be used to trigger a host of actions using Wavestore’s ‘Cause and Effect’ event rules engine. Streams with bounding box overlays are also available to ensure operators are instantly aware of detected persons or objects of interest within Wavestore’s WaveView client software.
WA-VADLOPT-1
Wavestore Deep Learning Object-Tracking and People-Tracking Analytics Single Channel Licence
WA-VADLOT-40
Wavestore DL Video Analytics Server for up to 40 Deep Learning Object-Tracking and People-Tracking Analytics Channels
.svg)
