A seamless video solution would include video intelligence at the camera (edge), video intelligence
shared by a “local mesh” of cameras to construct a local real time event unfolding, and intelligence
from local events further processed to construct a situation with greater detail and reliability,
with minimal human intervention.
The analytics requirements for residential, commercial & enterprise video solutions would vary.
The solutions are currently sought in the area of commercial and enterprise market segments.
In a commercial set up, like banking, shops, malls, with typically 4-100 cameras,
solutions are sought with edge video intelligence to identify abnormal or suspicious activity.
In an enterprise set-up, like airports, with 100 to 1000’s of cameras, solutions are sought to
identify the intent, context at the edge, and solutions are sought to track in real time any
unexpected event to raise alerts or awareness to security personnel
In many situations, the solutions that minimize specialized configurations and which eliminate
privacy concerns will be preferred. Solutions that integrate with other ubiquitous information
(like social media) to improve the accuracy and reliability of the data or to extend the
capabilities of the system, will be preferred.
The challenge therefore is to have analytics that have several of the below features:
Intelligent intrusion detection and classification of objects, indoor & outdoor
Automated people detection at parking entrance/exit
Vehicle / human detection – Further classification into subtypes will be given
extra consideration (e.g car vs truck, male vs female etc.)
People, vehicle, object counting, License plate recognition
Crowd density management
Static object (known as Left behind, abandoned baggage, left luggage)
Trajectory between cameras
People detection on a conveyor belt
People detection on railroad tracks or on gates
Meaningful data about the volume and direction of pedestrians,
bikes, motorbikes, cars, and large vehicles.
Which can easily be integrated with UTC’s intelligent software platforms via
APIs, Services or libraries.
The solution must be easily configured during deployment, with minimal setup
or training required from end users.
For systems that utilize off-line training of target objects, it is required that on-line
methods be available for improving or tuning the system when deployed.
And these are pulled together as needed, to find "a lost child wearing a blue jacket in mall"
or detecting "suspicious or abnormal activity at an airport" in real time.