Camera and AI technology to develop a 'smart' perimeter

The advantages of this technological combination, analyzed by Miguel Ballabriga, sales manager of Dallmeier Electronic Spain, result in an accurate solution, cost-effective and effective for securing the perimeter of any facility.

No perimeter solution looks like another, but almost every approach has two things in common: the combination and interaction of different technologies is complex to implement and manage. And, Moreover, most perimeter solutions suffer under the large number of false alarms that require a lot of staff in surveillance and decrease their reliability.

Developments in recent years promise remedy, as long as technologies are used correctly. In this sense, have highlighted two areas; on the one hand, analysis, where artificial intelligence (The) advances more and more in video analysis and promises a significant improvement in recognition accuracy. In this way, video analysis offers, for the first time, the possibility of partial 'automation' of verification in perimeter protection.

On the other hand, the area of camera systems: combined with modern analytics capabilities, certain camera technologies are increasingly taking on the tasks of detecting or, which is the same thing, the first event recognition.

Combination of technologies

Protecting a perimeter with conventional methods in a really effective way requires a combination of different technologies. There is no ideal solution, but, depending on the customer's environment, you have to opt for the corresponding complementary detection technologies, as tensioned wire systems, radar technology or accelerometers.

Most systems have high acquisition costs, Implementation (typically require infrastructure components such as wire and sensors at every meter of the perimeter) And, At last, operating. The fact that no system alone is sufficient multiplies expenditures again.

Then, wouldn't it be obvious that camera systems, that often already exist for event confirmation, they could take on all the task of detecting or, at least, a part of it? This is possible if the camera solution is designed to do this.

The system has to have the basic functions of VCA (video content analys) triggering an alarm as soon as an intrusion is detected, For example, intrusion detection or line crossing.

However, there are reasons for the continued use of camera systems in perimeter protection only for verification and not for event recognition. A common reason is that it is not possible to plan the quality of resolution: an analysis can only be as good as the quality of the data to be analyzed.

So, users should attach great importance to a provider offering opportunities to specify, with little effort, the exact minimum resolution density across the perimeter to be captured for successful recognition of an object or person.

This is defined in DIN EN 62676-4, globally valid, and it's set -- depending on the scenario- Between 125 And 250 pixels per meter for recognition and identification, Respectively, of a person. The precise definition, already in planning, is a basic requirement for analysis, preservation of evidence and additional fine analysis using AI media.

Proper lighting

Of similar importance is the choice of proper lighting: if it is possible to use white light, a powerful camera system is usually sufficient for rainy day and night situations. Good systems are able to ensure sufficient recognition of objects with active infrared spotlights, as well as identifying people, even under poor light conditions.

For poor light or weather conditions, it is advisable to complement the solutions with thermal cameras. These are harmless in terms of data protection, while they do not allow any identification of people and preservation of evidence, so you have to work with connectable spot white light or surveillance staff.

PTZ camera inconveniences in verification

Also in verification you can avoid many mistakes by choosing the right technology. Modern 'multifocal sensors' cameras, like Dallmeier's Panomera, are distinguished from the classic 'single sensor camera plus PTZ camera' combination by their constant minimum resolution densities, even across long journeys, requiring fewer cameras.

Added to this is the ability to open zoom views at will for verification, keeping complex situations such as the intrusion of several people under control as well. Unlike conventional solutions, you can zoom in on both live and playback images with high resolution quality, allowing you to capture any area in detail also at a later date.

The: false alarm quota close to zero

In addition to a minimum number of systems for perimeter protection, the greatest optimization potential is in partial automation of tasks that, Currently, still require a lot of staff and time.

A central role is played in this context by artificial intelligence. With reasonable skepticism, already and especially in perimeter protection, concrete solutions that promise a huge benefit to the customer in staff costs and, with it, total operating costs.

Dallmeier's solution addresses the pre-alarms generated by the classic VCA - including the usual 'sources of error', like moving branches, rapid changes in light conditions- in a second level of neural network analysis.

In the 'AI engine' a certain probability of recognition is defined, and only alarms exceeding this value are transmitted to the operations center for 'human' verification.

The results can be seen: experience at early customer facilities shows a reduction in false alarms to virtually zero and, therefore, the obvious decrease in the effort to check or track alarms.

In this way, users achieve huge cost savings, because they can protect with the same staff much larger perimeters and, In addition, objective safety increases by significantly higher attention from operators. After all,, know now that behind an alarm warning there, with a very high probability, a relevant incident rather than another false alarm.

Miguel Ballabriga

Sales manager Dallmeier Electronic Australia

 

 

 

 

Other articles on Dallmeier