The pose detector reacts to different non-standard behavior scenarios. At the same time, the artificial intelligence (AI) that controls it is capable of simultaneously working with hundreds of video sources without direct human involvement. This takes security at the facility where such a system is installed to a new level.
– The university, whose educational buildings are located in various microdistricts of Nalchik, has a digital access control system (ACS), which provides anti-terrorist protection for university facilities. All entrance groups are guarded, they are equipped with a system of direct communication with the National Guard, – KBGU rector Yuri Altudov told the “RG” correspondent.
All data from the cameras is transmitted 24 hours a day to the situation center, where specialists are on duty. They monitor what is happening on campus and respond to incoming signals from security posts and other security systems.
They test the AI system for several months, while the artificial intelligence continues to learn already in a specific area. The work is carried out by specialists from the Department of Information Technology and Information Security of the Institute of Artificial Intelligence and Digital Technologies and from the Center for Security Monitoring and Management of the Department of Infrastructure Engineering and Information Technology. According to Albert Shapsigov, director of the institute, 520 IP cameras are currently operating at the university.
– They are installed in educational buildings, dormitory buildings and on the adjacent territory. The artificial intelligence systems that the university specialists are testing were developed by our partner, the ITV Group company, and allow computer analysis of events that fall within the field of view of video cameras, Shapsigov noted.
The neural network analyzes the behavior of people 24/7. The detectors are activated by certain potentially dangerous situations, such as “man shooting”, “man lying”, “hands up”. Such a situation may be a sign of an armed attack or robbery: one or more people stretch out their arms in front of them, as if holding a weapon, while others raise their hands above their heads. In this case, the system will automatically notify the security forces or the facility’s security service about a possible emergency.
If a person is lying on the floor or on the ground, it is a sign that they probably need help. A quick response to a fall in public places, buildings and industrial facilities will help reduce the risk of negative consequences. To which postures and scenarios the system responds is decided by those who operate it. The first test results at KBSU are encouraging: the detectors correctly recognize situations and give an alarm in a timely manner.
As Professor Tatyana Khashirova, Head of the Department of Information Technology and Information Security at KBSU, said, in addition to improving the security of the university, the use of AI helps the teaching of students.
– Detectors are only effective if they are used correctly and the parameters are optimally set. Direct communication with the software developer helps us with this,” Khashirova explained.
After the tests, KBSU will decide where and how to use the system in the future. Most likely, in order to improve the general level of university security, these will be the most vulnerable and significant areas of the territory. .
Dmitry Budkov, programmer:
– The described technology exists more than six years. At the moment, there are already quite a few different architectures that work on the same principles. The system keeps track of 10-20 so-called “key points”: head, eyes, arms, legs, depending on the settings. The position of these points is compared with each other, and the system concludes that the person is sitting, immobile for a long time, with his hands in an unnatural position.
In any case, artificial intelligence technologies do not provide 100 percent reliability. However, the fact that many companies use them in practice speaks in favor of their use. In retail – to analyze the behavior of buyers or warehouse workers, in construction sites – the behavior of builders for timely response to emergencies and their prevention. The scope is quite wide.
There are no restrictions on the number of people in the framework, but data processing depends on the hardware capabilities of the equipment on which the system is implemented. In operation it is not difficult if it is supplied as a finished product, but in any case it requires prior configuration by IT specialists.