Smart Office
A perfect office is one where you feel the employee-friendly atmosphere everywhere from the security desk in the entrance area to the back office operations.
In reality, however, it can be quite difficult to combine openness and security. Imagine a corporate office visited by tens and hundreds of people every day: regular employees and outsourcers, guests, couriers and drivers.
In reality, however, it can be quite difficult to combine openness and security. Imagine a corporate office visited by tens and hundreds of people every day: regular employees and outsourcers, guests, couriers and drivers.
IN ITS SEARCH FOR A COMPROMISE BETWEEN OFFICE OPENNESS AND SECURITY, THE COMPANY DECIDED TO USE ARTIFICIAL INTELLIGENCE AND CONTACTED CENTER2M
The company decided to implement the "Gateless ACS" concept: a special security system that can identify employees and monitor the flow of visitors while maintaining the atmosphere of openness and accessibility.
It is impossible to create such a system only relying on the human factor. Not even the most attentive security guard can memorize hundreds of faces, especially since many employees go on business trips and may not appear at the Moscow office for several months.
It is impossible to create such a system only relying on the human factor. Not even the most attentive security guard can memorize hundreds of faces, especially since many employees go on business trips and may not appear at the Moscow office for several months.
Request
Solution
If the system e.g. has to identify a person quickly and accurately, the image must have the following parameters:
• Face width: no less than 250 pixels;
• Space between the eyes on the photo: 150 pixels;
• Distance between the eyes: at least 40% of the total face-width.
It is easy to have such an image if persons look straight into the camera and much more difficult if they decide to swerve a little or take a step to the left ending up in a blind spot. That is why it was very important to install all the cameras correctly and ensure the maximum view of the scanned perimeter.
• Face width: no less than 250 pixels;
• Space between the eyes on the photo: 150 pixels;
• Distance between the eyes: at least 40% of the total face-width.
It is easy to have such an image if persons look straight into the camera and much more difficult if they decide to swerve a little or take a step to the left ending up in a blind spot. That is why it was very important to install all the cameras correctly and ensure the maximum view of the scanned perimeter.
At first glance, the scenario was simple. The Security Service inputs employees' photographs into the database. The CenterVision system analyzes the flow of visitors and compares the faces to the Security Service database
The first step was to install cameras and connect them to the server that had the CenterVision system with a built-in biometric face recognition scenario. There were no difficulties with the installation: a camera with 2 MP resolution and recording speed of 25 frames per second was enough for the scenario. We chose the HIKVIZION model that suited very well both by price and by quality.
We used the Postgres ACS, which is a very popular and easy to use system. The Linux platform worked quite well with CenterVision. HP server with Nvidia TESLA T4 ensured the best scalability/price ratio. Excellent tech support from the manufacturer was a bonus feature.
We used the Postgres ACS, which is a very popular and easy to use system. The Linux platform worked quite well with CenterVision. HP server with Nvidia TESLA T4 ensured the best scalability/price ratio. Excellent tech support from the manufacturer was a bonus feature.
Create a convenient and unobtrusive security system that can automatically monitor the flow of visitors
Train the system to recognize guests, identify insiders and alarm the security service when strangers appear
Install cameras equipped with the CenterVision system that has a biometric face recognition function
However, the devil lies in the detail and details often determine whether the system works, how often it creates mistakes and whether it will mistake an accountant who decided to grow a beard for a stranger
Result
The result of Center2M work was a system of smart video surveillance that identifies visitors in real time
Up to 98%* recognition accuracy
Starting at 300 employees daily
Yet another pitfall in biometric recognition are emotions. Everyone is calm and serious on ID photos, but we come to work in very different moods sometimes, with a frown on our faces or with an expression of subliminal bliss from meeting with colleagues. Emotions that are too vivid may interfere with clear face recognition; that is why CenterVision can measure the visitors' emotional state on a scale from 0 to 100%. Company employees are aware of this function; it has become a type of entertainment for them. It is fun starting a day knowing that you are 15% calmer than your colleague from the next department.
In action movies, villains deal with biometric cameras by turning their hat forward and raising the collar. This only shows that movies use the old systems. New systems, such as CenterVision, easily identify a person with any type of a "disguise": eyeglasses, a wig or a hat.
«beards and mustaches may become a more difficult obstacle, but here the ball passes to the Security Service: if they notice significant changes in the employee's appearance, they can simply add another "bearded" photo of the person to the database.
«beards and mustaches may become a more difficult obstacle, but here the ball passes to the Security Service: if they notice significant changes in the employee's appearance, they can simply add another "bearded" photo of the person to the database.
CENTERVISION IS CAPABLE OF MEASURING THE VISITORS' EMOTIONAL STATE
CENTERVISION EASILY IDENTIFIES A PERSON WITH ANY TYPE OF "DISGUISE": EYEGLASSES, A WIG OR A HAT
Of course, there are better ways to cheat biometric recognition cameras. For example, there are spoofing attacks, when illegal intruders try to enter the premises using a photo of a real employee. CenterVision blocks such attacks using the special anti-spoofing system.
An attempt to pass the monitoring area using someone else's photo (spoofing attack)
*If the conditions, such as correct installation of cameras and required lighting, are met, system errors are reduced to a bare minimum
Perspectives:
Further plans to grow the system include:
Close integration with the client's database
On-demand reporting
Adding license plate recognition