printer friendly version

Finding people in real time

Rapidly locating a person of interest is one of the most important tasks for every operator in the control room. Whether it is working in real-time to reunite a lost child with their parents, aiding in the apprehension of a suspect, or working with the authorities post incident, every second counts. Improving capabilities in this area has resulted in one of the most exciting developments, in what is being referred to as next-generation real-time video analytics.

Erez Goldstein, senior product marketing manager at Qognify.

For estates that have large numbers of surveillance cameras distributed over wide areas and multiple sites (such as airports and other mass transit hubs, hospitals, sports events, city centres and government facilities), finding person/s of interest has traditionally been a resource draining and time intensive operation, often lasting hours. Furthermore, major incident investigation could result in days, weeks or even months of manually trawling through masses of CCTV footage.

What is more, while pinpointing the whereabouts of the person is paramount, if a ‘live’ situation presents a possible threat (such as the suggested presence of a known suspected terrorist), this information alone isn’t enough. You need to know when and where they entered the estate, the precise route they took, who they were with and what they did before arriving at their current destination.

To address this precise issue, the latest video analytics (such as Suspect Search from Qognify) have been proven to help reduce search time from hours to minutes and even seconds. To best explain how it works in practice here is an example.

Search on description

The information desk is notified by a concerned traveller that they have seen someone behaving very suspiciously in the land-side shopping concourse of an international airport. The control room is notified about the incident and is given a description of the person of interest. The operator uses this information to quickly make an avatar that matches the description – a white male, medium build, blonde hair, wearing blue jeans, a green hooded sweatshirt and a black backpack.

Every relevant camera on the network is then searched to find all instances where person/s matching the description appear on camera. Three to five seconds later the operator is presented with a selection of images from which he fine-tunes his search (more than 95% of irrelevant images have already been filtered out by the algorithm). He identifies the person in question and now uses a still image from the camera as his reference photo instead of the avatar, re-iterating the search, which now provides even better results for getting a match to the right person.

The rapid speed of the search is made possible because the system is capturing, indexing and storing data in real time, from every camera. What is more, (in the case of Suspect Search) the system is camera agnostic and has basic minimal requirements only for image quality and size (a plus point over facial recognition technology).

Historical routes

With the person identified, the operator is then presented with a clickable map detailing the route, locations and times that the person is visible to the camera, as well as the most updated video feed showing where the person was last (or even currently). The operator now tracks the movements of the suspect, but crucially he no longer has the backpack and the priority is to locate it immediately. Reviewing the footage, the person is first seen talking to a person he did not arrive with and he hands over the bag. Now the operator uses the system to search for the second suspect (whilst continuing to monitor the first suspect) who is seen entering a café and leaving without the bag.

Now monitoring all three live-feeds (the two suspects and the nearest camera to the bag) the operator is guided step-by-step by the situation management system, following the pre-determined protocol to co-ordinate a response. He dispatches one security team to the café and triggers the evacuation of the surrounding area to an agreed boundary, at the same time the on-site police are sent images of the suspects, along with their exact location, to ensure they are apprehended for questioning.

Not only has the system been used to locate the suspects and assess the severity of the situation, all of the time-stamped video footage can be packaged for review by police as well, for evidence submission.

For more information about real-time video analytics and Suspect Search visit www.qognify.com

Share this article:

Further reading: