The next bis thing - KI Nutzung im Sicherheitsbereich

Even though robotics is in our name, mobile, autonomous and very versatile assistants, supporters and colleagues are only ever one part, the hardware part, when we talk about innovations and modern solutions for the security industry.

The use of AI, which fulfils a wide range of tasks in the security sector, solves them faster, more efficiently, more securely or makes them possible in the first place, is even more important and offers far greater possibilities.

How important is AI in the security sector?

Artificial intelligence is becoming active in many areas, is constantly learning and can therefore make more accurate statements, make recommendations and create facts. But it is not just AI that is learning; we are also getting better and better at interacting with each other, using artificial intelligence as a tool as a matter of course and opening up more and more possibilities.

We are currently seeing three main areas in which AI is bringing about decisive changes and massive quality improvements:

Video analysis / image evaluation

AI systems, whose routines are trained in advance and continue to learn from the data collected, are being used extensively here. Both on board security robots, as well as in video management systems, of course our platform solution ACUDA – which bundles data sources of all kinds – and finally in modern emergency call & service control centres, these systems help to analyse static image material and video streams of all kinds in real time and search for patterns.

In principle, people, vehicles, objects, obstacles, etc. can be recognised and classified. In practice, this makes it clear if, when + how many of these elements are in a defined area, moving in it, crossing it, etc. Depending on the requirements, the AI pays attention to vehicles or even people, but can also react to parked goods / merchandise, for example.

If the area of restricted areas/zones is violated, the AI system recognises this with ease and reports the incident, including all relevant data, to designated recipients such as an NSL. As this is live data, sensory tracking of intruders is also no problem, i.e. security technology integrated into the system such as cameras, robots or even security personnel always have the current location, vector and type of intruder available.

As far as is compatible with data protection law and defined as part of a security concept, vehicles can be individually recognised based on their number plate and appearance, which makes automatic and logged access possible. Movement patterns, facial analyses or the sensory recording of biometric data also make it possible to recognise people not as one person, but as THE individual and to grant or deny them access depending on the rights management of an access control system.

It should also be noted that objects moving from / in certain directions and at certain speeds can be recognised as potential dangers and reacted to even before they leave the monitored area and reach their destination. This type of early hazard detection and the resulting action is unique.

The automatic, self-learning AI image analysis tailored to your task can therefore make a decisive contribution to automating classic security tasks such as property protection, perimeter protection and access management, with human decision-makers only having to intervene if necessary.

However, the possibilities are not limited to keeping an eye on access itself, but also, for example, to fulfil occupational health and safety tasks. AI models can be individually trained to recognise specific features and use these as triggers. In our example, we developed the AI model in such a way that it is orientated towards wearing or not wearing a high-visibility waistcoat. If this is correctly attached to the upper body, i.e. clearly visible on a person, they are considered authorised. If this waistcoat is missing or is only in the hand or under a jacket, the person is considered unauthorised. In this customer-specific case, each authorised person is also automatically rendered unrecognisable in compliance with the GDPR and no longer identifiable, a process in which an AI also played its part. In response to the registration of an unauthorised person, the robot on site could address them and expel them from the premises without the need for a human colleague to come on site.

During regular maintenance, inspections or for the documentation of safety-related requirements, it is often necessary to check the presence of objects (fire extinguishers, signage, etc.), their condition and functionality and to make them available digitally – sometimes in real time. AI can also provide decisive impetus in this area and the reading of scales, measuring instruments, signs, markings and seals becomes a workflow that, once implemented, takes place automatically from the perspective of a fixed camera, during a robot tour or when flown over by a drone. Deviations are registered, documented and reported, and potential problems with systems are recognised before they can have avoidable consequences.

Thermal cameras can also provide important services with AI image analysis, recognise temperature fluctuations, locate heat sources, locate and report people, animals and vehicles at night and in extremely poor visibility based on their heat signature.

Even if it is mostly about cameras, the topic is not limited. Acoustic data, radiation values, gas density and much more can also be analysed. It is all a question of the requirements, technically feasible and is already being realised by us on behalf of customers.


Big-Data Analyse

Networked security technology, and this also includes – in an expanded sense – access systems with RIFD chips or GPS location data from machines etc., generates a large amount of data: Movements, activations, opening/closing, audio signals, video streams, images, position reports etc. so that it is theoretically clear when what happens where, where it is located and from the time periods of activations, their frequency, sequence and combination with other recorded signals, conclusions can be drawn about the quality and efficiency of operational processes, the level of security achieved and potential weak points.

Linking and evaluating such information in a meaningful way, writing algorithms that allow analyses in the first place and ultimately creating a system that develops useful recommendations from rules, experience, feedback and live data is a very complex task. However, the data is available in principle and, according to our vision, could also be used with greater added value in particularly vulnerable environments, KRITIS areas and, of course, completely digitalised security systems.

This is still largely a dream of the future, but it is up to us to derive scenarios and possibilities from this that could represent the next level of AI-supported security.

Relevant data from incidents, sensor messages, alarms, etc. are already being used to preserve evidence thanks to automatic storage and allocation, so that incidents are almost completely documented and traceable. Instead of real AI, however, this “only” requires a well thought-out management system.


Routing / Movement

In order for robots to be able to move autonomously and cope with any surface, slopes, changing weather conditions, sudden obstacles or particularly challenging things such as stairs, they have to learn to drive and walk. This learning phase, which every child goes through, consists of countless repetitions and lots of variety. Trial and error enables the system to make increasingly stable, efficient and safe movements.

So before a four-legged robot like SPOT can smoothly master a parkour, it has to slide, fall and get stuck countless times. Interestingly, not all training is carried out in real life, because this is precisely where AI can create virtual training sessions, constantly adapting, analysing and redefining these “tests”. And, another important point, dozens or hundreds of simulations can be run in parallel, greatly accelerating the pace of learning.

AI also helps with routing, i.e. orientation in its environment, reliably following a route and determining its own position. As a result, an ARGUS, for example, is very faithful to its tracks, i.e. it turns lap after lap almost to the centimetre in its own tracks and can also cross narrow areas thanks to the precision it has achieved.

Stronger togehter: KI + Robotics + Humans

It is crucial not to look at AI in safety technology in isolation, but to realise its possibilities and strengths and incorporate them into safety concepts. Our example with the high-visibility waistcoats shows the added value and new features that are suddenly conceivable. It’s not just about detecting people in the first place, but also using this data in parallel for other tasks, in this case to check whether they are wearing a safety waistcoat and, if the result is positive, to finally anonymise it without any manual intervention, additional clicks, etc.

In short, a key technology such as AI can provide support, make things easier, enable new use cases and, of course, significantly improve the quality of work. So let’s use “the next big thing”, it’s worth it!


#KI #Sicherheitstechnik #Innovation #Robotics #Digitalisierung #SafetyTech #BigDataAnalyse #TechRevolution #SmartSecurity #NextGenSecurity #SecureFuture #HumanAIHarmony #DataDrivenSecurity


Michael Engel

Phone: +49 341 2569 3369

Contact us!