Video Based Artificial Intelligence: How Smart Algorithms are a Game Changer in Video Surveillance and Access Control
To many, AI-powered Surveillance may sound like the plot of a futuristic Hollywood movie. The technology, however, is no longer science fiction. Video-based Artificial Intelligence systems can recognize objects, humans, and biological characteristics like faces, the iris or even emotions and behaviour. They can be customized to specific recognition requirements like a-normal behaviour in crowds, unaccompanied children or emergencies and can be embedded into any surveillance landscape. They can even recognize specific characteristics of objects and detect deviations or manipulations. AI technology allows law enforcement and first responders to quickly act on information during an emergency. In the near future AI powered surveillance will play an important role in crime prevention as well as crime analysis.
But how exactly are AI surveillance systems so dramatically different from existing video or surveillance approaches?
Well, essentially it is the speed and the accuracy at which they operate. While traditional surveillance systems require humans looking out for suspicious behaviour on CCTV screens or trawling through hours of recorded footage for a single perpetrator or a specific moment, AI can bypass these limitations. The technology searches for specific traits, patterns or anomalies in real-time and at a rate that humans could never meet.
In a time where many advocate better monitoring and more surveillance cameras, more and more video footage is being produced. Human capacity limitations and abilities can be restrictive while smart algorithms will never get tired or bored, they can work 24/7 and they will never miss a detail. This makes artificial intelligence ideal for identifying threats, emergencies or, even, specific people at a moment´s notice.
Protecting Large Scale Events, Publicly Accessible Space, Infrastructure
With growing and changing threats to normal life and activities as witnessed across European cities and indeed around the globe, the demand for more effective surveillance gets louder and more essential.
Ensuring the safety of individuals attending large scale events gets more challenging and complex. Organizers of such events have to take a much wider range and unprecedented scope of terrorist methodologies into account. These include hostile actors using firearms and vehicles as weapons but also abandoned objects like suitcases placed in an airport lobby which might hide a bomb. Even alterations to objects may result in substantial damage to individuals: As an example, Cash trapping, which, like it sounds, blocks an ATM’s cash dispenser shutter so that bills cannot be presented, is a constant threat. Smart algorithms can improve ATM hardware security by instantly detecting manipulations on the ATM via a standard surveillance camera mounted in the ATM location.
Generally speaking, security controllers should understand risks holistically with regards to both crowds, locations and infrastructure, as well as possible perpetrators. With AI systems they can easily recognize patterns, deviations and outliers. They can foresee activities and, most importantly identify a-normal behaviour.
An example: AI-powered cameras at events can radically improve security: Unusual activity (pickpockets, riots, assaults, etc.) or trespassing can be automatically detected using behaviour patterns, allowing a quicker response time when things escalate to a crime or emergency. Even better: AI systems can play an important role in the prevention of crimes ahead of an event. With the pre-surveillance of a location, smart technology can recognize persons with suspicious or unusual behaviour, potentially checking out the location in order to prepare or plan their crimes. They can easily differentiate a simple stroller by from a potential perpetrator. Even better: As gait, like the iris or a finger print, is a unique biometrical identifier for each human being, leading AI solutions can even highlight suspicious persons by the way they walk.
Access Control
Also for any kind of access control AI can provide effective and smart solutions. For buildings, facial recognition is becoming the most common and critical form of recognition. Think of how today you walk through automated gates at passport control or entry in to you apartment block. Usually a photo in a database is matched against the video recording of the person requesting access. Once the AI systems recognize specific persons, doors or turnstiles/entrances open. In case an unauthorized person wants to enter, a specific signal or alarm notifies the surveillance person. Hence, the same principle: Notifications are autonomous and in real-time. Humans only come into play when a security breach happens which proves to be very effective.
More sophisticated systems have additional advantages. As often privacy is a concern, the encryption of personal data (photos, videos) is essential. Also, there are huge quality differences when it comes to recognition capabilities. Whilst the less expensive systems can only operate in good light conditions and request the person to show his/her face full front into the camera, the more advanced systems can work reliably under adverse conditions and with the face even partially obscured by, say, sunglasses or a hat.
Whenever building security has to meet highest standards, face recognition can also be combined with complementary recognition systems like motion analysis. Modern access control systems can be customized to any specific requirement. Thanks to open APIs they can be embedded into any IT environment. These intelligent solutions will soon be found in airports, in vehicles of any kind or in locations where highest security requirements must be met. Also, they might soon be a standard method in companies, simply when clocking in or out. They will make access easier and more convenient and they will play an important role in reducing location specific risks.
More Use Cases: Counterterrorism, Border Security, Activity Monitoring
Whilst a comprehensive strategy to counter crime and terrorism must incorporate many elements including diplomatic, economic, financial, law enforcement, and police instruments, AI powered recognition will certainly play a critical supportive role both in crime prevention as well as in after-event analysis. It can dramatically speed up identifying perpetrators and make crime solving easier.
- Detect faces in real-time by running them against a police/government watch list. Already known suspects will automatically be identified
- Existing or historic video material can be leveraged for more accurate and in-depth analysis
- Identify unknown repeat offenders by matching video footage with faces and motions.
- Automate the detection of unauthorised border crossings by monitoring activities (motion recognition) also with ultra-red cameras.
- Analyse mental or physical ability to operate a machine or drive a vehicle: Detect drowsiness and intoxication. Detect unwanted behaviour, e.g. answering the phone while driving a vehicle carrying hazardous cargoes
No doubt, AI-powered video recognition will revolutionize the way in which we make locations and events more secureThere are numerous use cases for which this smart new technology can make a tremendous difference in terms of efficiency and safety. It will play an increasingly important role in every-day life, and it will become standard technology – much sooner than most of us imagine.
About DeepEyes
Disruptive video-based recognition for requirements until now considered out of reach.
DeepEyes is an A.I. technology based on video-generated data, which delivers highly accurate results in real time without a need for cloud access or traditional grid computing. The algorithm operates autonomously on standard hardware.
With DeepEyes it is easy to build secure predictive models for all high-performance visual analytics requirements in any industry.
For further information:
