Video Analytics for Crowded and Complex Scenes
In November 2011 iOmniscient, pioneers of intelligent video analytics, celebrated their tenth year of trading. The company based in Sydney, Australia, specializes in providing video analytics for crowded and complex scenes and offers a range of comprehensive solutions for over 30 industries. Its latest product is "Face Recognition in a Crowd".
This particularly advanced system can recognize people at a distance in an uncontrolled environment. Requiring a resolution of only 22 pixels between the eyes, it can outperform humans and all other current face recognition systems which require much higher resolutions and can only operate in controlled environments. iOmniscient's "Face Recognition in a Crowd" has immense practical value in a number of situations described below.
Firstly you can identify people from a "black list". They may be known criminals or terrorists who are attempting to enter a country incognito. Other examples would be the provision to shopping mall security of a black list of hundreds of known shoplifters, or to casino security of a list of crooks.
With the new Face Recognition system, a security officer can check out suspects encountered in his or her daily work through their smart phone. One of the major issues in policing, which organised criminals can and do exploit, is the lack of communication between different forces. A face recognition system like this could provide a genuine, instantly accessible national criminal database to overcome the problem.
A huge database would require high computing power, however, with iOmniscient's Face Recognition a database of 250, 000 faces can be handled on a simple server.
Using Dynamic Databases
Secondly, "Face Recognition in a Crowd" can also be applied with a dynamic database. An example would be passengers walking down the airbridge whose faces are detected. Then, if they turn up at immigration claiming refugee status and refuse to say where they have come from, the system can check on the database of passengers disembarking from different aircraft, to ascertain the flight on which each individual has arrived.
Thirdly, the system can improve customer service.
It can produce a "white list" of VIPs or most loyal customers whose faces are in the database. Then they can be immediately recognized in say a hotel reception or casino and afforded privileged, red carpet treatment.
A fourth application is driver matching. Instances might include a truck driver for a valuable cargo, or an authorised driver for a luxury motor car on hotel or club premises.
"Big Data" Using Face Recognition
A fifth, most beneficial method of deploying Face Recognition is for managing large amounts of unstructured information, which is now increasingly known as "Big Data". One example of Face Recognition used in this way is where a face is recognized at the start of a queue. When the same face is seen at the end of the queue, the system can provide management with information on the average time it takes for people to pass through that queue. Such a system can help manage complex queues with multiple entry and exit points.
Degree of Accuracy
In an uncontrolled environment, handling variable face angles, the "Face Recognition in a Crowd" system will typically be 70% accurate, because three out of ten faces may be covered up or directed away from the camera. So the clever bit consists of getting people to look up at a concealed camera - positioned for example at the top of an escalator, or next to a sign on a concourse such as "Arrivals - Turn Left" to catch their faces. In those scenarios the success rate of facial recognition can be higher.
What of the Future?
Application areas for "Face Recognition in a Crowd" will be seen to extend beyond the more obvious current areas such as airports and ports, banking and retail, border and city surveillance, crowd management, hospitals, hotels, manufacturing, police and prisons, transportation and utilities.
Secondly, more and more developed and developing nations alike will want to take on this new technology to address their security concerns. Thirdly, computer memory and power will increase, facilitating larger databases which will handle millions of faces.
Finally, in a future release, people's faces may be distinguished according to age, gender, skin colour and ethnic type, potentially fulfilling a host of intelligence gathering operations. In addition to national security and law and order, applications for that may include market research, retail footfall, passenger counting, employment, and health and safety. Privacy laws differ from country to country and therefore, the iOmniscient Face Recognition in a Crowd system can be adapted to suit the laws of the country.