As cameras have become smarter, technology can now detect objects and faces and identify many unique characteristics that differentiate them.
Computer vision systems that utilize artificial intelligence (AI) and face detection are fairly well understood in our industry. The concept of persons of interest being on a watch list is well known. However, there are other use cases that are worth exploring as we imagine how this technology can be leveraged to help protect individuals and businesses.
Most businesses that have surveillance video have a policy for retaining that video for a period of anywhere from 30, 60 to 90 days depending on the type of business. As cameras have become smarter, we can now detect objects and faces and identify many unique characteristics that differentiate them. This additional data can also be stored alongside the video. As VMS systems mature, they can read this additional data and use it for highly efficient and targeted forensic search tasks. For example, “show me all red cars between 1 and 3 p.m.”
With face recognition, when a face is captured without a match, it is possible to still retain that anonymous face data for the same retention period as the video. These faces are not matched to anything (assuming they are not on a watchlist), and they are not identified in any way. They are simply part of the data extracted and retained from the video. If there is a situation that warrants further investigation because of the self-interest of the individual or establishment, the data that is already available on file may give us a starting point.
Level the Playing Field
Computer vision makes it possible to investigate crimes that previously did not have sufficient information to start on because the cost and manpower required to mount an investigation was too great. While serious crimes might justify entire departments of professionals pouring through every video feed in an area, smaller crimes against ordinary citizens rarely, if ever, warrant the effort.
Computer vision technology could enable us to bring justice and perhaps restitution to those victims who in the past, were never going see anything be done on their behalf. Businesses can better protect themselves and their customers. Through the use of computer vision and face recognition specifically, we can now more easily say, “Here is a person. Show me all persons that have appeared in front of a camera with this person within X number of seconds.”
In this way, we’ve just achieved what a large team of professional investigators might take days to achieve in just a few minutes. Law enforcement can bring their other resources to bear to put names to the anonymous faces as is necessary.
Casinos and bars continually worry about over-serving alcohol to patrons. Although laws vary by state, “dram shop” laws can hold establishments responsible for damages caused by knowingly overserving patrons. Being able to keep an archive of all faces that come through an establishment can be invaluable for businesses wishing to protect themselves and their patrons from wrongful lawsuits.
By keeping anonymous face data for a given retention period, security professionals could find all the points in time when that individual was seen on video, especially when they arrived on the property. Did they arrive on the property, according to the video, looking completely sober? Who did they arrive on the property with? Reducing liability by having comprehensive evidence can be extremely valuable.
In the age of COVID-19, how can we potentially identify individuals that may have come in contact with someone who has tested positive? Contact tracing is ramping up as we continue efforts to defeat the virus, but it requires a lot of manpower to successfully carry out. Can we more efficiently identify who may be high priority for secondary screening?
Once an infected individual tests positive, can we find all the places where they went as they were walking through the airport, and either pay greater attention to the surfaces and areas that staff and fellow travelers shared so as to get the appropriate notifications out? How can our existing surveillance technology help us in that regard?
That’s one area where a post-event investigative workflow supported by face recognition could potentially be of use. The technology is performing the same task a group of people could do, but it’s doing it much more efficiently at a significantly lower cost.
There are more cameras out there than ever before. Adding still more is not necessarily the answer because we don’t have the manpower to monitor and review all those streams. By using computer vision, people don’t have to spend hours watching and monitoring, and reviewing video feeds. You can skip over the vast majority of people that are not involved in an incident, keeping their privacy intact while letting the technology identify the most likely individuals that warrant an additional human review.
Making Cameras Smarter
Whether it’s tracking contact between virus carriers or protecting people and businesses, the efficiencies delivered by computer vision technology can transform our ability to efficiently investigate events that previously would have been ignored. To advance the use of the technology, it’s critical to understand that people are not being identified.
The data that is stored is anonymous (outside of any watch list or opt-in individuals) and it stays anonymous until there is a need to discover the truth about an event that warrants interaction. The cameras are probably in place already, so it’s just about making them smarter so they can assist us in protecting each other.
Eric Hess is Senior Director of Product Management for SAFR, a facial recognition platform for live video intelligence, from RealNetworks.