Step 1
The camera is connected as a source
A workshop camera is added to the account as a video source, an archive video — as a file. Sources are combined into groups: “Workshop 3”, “Warehouse”, “Gates”.
AI video analytics for manufacturing, shops and site security
The NeuroVision video analytics system works with the video from your cameras and with archive recordings: it helps to flag events by your rules, count visitors and find faces from your lists. You get an alert and a report, not hours of footage. Which scenarios to switch on — manufacturing, a shop or site security — we decide together in a pilot on your recordings. Faces are matched only against lists of people who have given consent; the other scenarios work without establishing identity.
NeuroVision video analytics
Cameras and archive videos Lists of people and alerts Faces — only with consent
A person makes the decision on an event; faces are matched against lists of people who have given consent
Intelligent video analytics: the camera notices the event itself
You already have cameras — you need analytics
We connect your cameras and archive recordings as sources. We choose the camera requirements for the task and check them in the pilot.
- 01
Events instead of hours of footage
An ordinary video surveillance system writes an archive, and it is opened when something has already happened. AI video analytics processes the video straight away and flags what matters by your rules — for example, a person in a restricted area or a crowd at the tills. The log holds a frame and a time, not eight hours of footage. We confirm the set of rules in the pilot.
- 02
The decision stays with a person
Video analytics does not punish and does not open the gate by itself. It sends the event to an occupational safety engineer, a shop manager or a security guard, and the person decides what to do. You discuss the rules, zones and thresholds with us and change them as you go.
- Cameras and archive videos as sources
- Lists for faces — only with consent
- A person decides what to do about an event
What the NeuroVision video analytics system can do
A video analytics platform with five scenarios. They are switched on one by one — for the task and for the cameras already installed at the site.
PPE control by camera
A scenario for workshops and sites: flag that a person in a zone has no hard hat, vest, mask or goggles where they are mandatory. Which PPE and in which zones is taken into account, we define with you in the pilot. The system does not establish who exactly broke the rule: the event holds a frame, a zone and a time.
Counting visitors and queues
A scenario for a shop: how many people came in, how many are standing at the tills, how much time they spend at a display. The count is anonymised: people are counted, not recognised. We refine the set of metrics in the pilot.
Face recognition in video surveillance
Faces in video and in the archive are matched against your lists: employees, guests, customers who have given consent. The source is a camera, a photo or an archive video; you can set a minimum face size and an acceptable head turn. Gender, age and emotions are calculated in anonymised form, for flow analytics; for age estimation by face — see the face recognition page.
Zones, lines and vehicles
Scenarios for security: entering a restricted zone, crossing a fence line, a person in the frame at night, a vehicle at the gate. The zone is drawn on the camera frame together with the rule and the time; reading vehicle licence plates is discussed for the project.
Searching the video archive
Archive videos are connected as a source. Instead of manual viewing, you search by events and by people from your lists. A face search works from the photo of a person on the list; how a search in your own database works is described on the page on face search.
How AI video analytics works: from frame to alert
An example — a metal structures workshop where a hard hat and a high-visibility vest are mandatory in the overhead crane zone.
Step 2
A zone and a rule are set on the frame
A rule is set for the crane zone — for example, “hard hat and vest are mandatory from 7:00 to 20:00”. Another camera has its own rule. How exactly zones and schedules are set, we clarify in the pilot.
Step 3
The neural network analyses the frames
The system finds people in the zone and checks them against the rule. An event is recorded if the violation lasts for the set time — a stray frame does not turn into a false alarm. We configure the duration and the threshold in the pilot.
Step 4
Alert and report
The responsible person receives the event with a frame and a time and marks it “resolved” or “false”. For faces from lists, alerts arrive on a match and on an error. We clarify the alert channels and the form of the report in the pilot.
Video analytics for manufacturing: PPE control and occupational safety
The employer is obliged not only to issue protective equipment but also to organise control over its correct use. Occupational safety with video analytics means PPE control by camera in every shift, not only on a scheduled walk-round. Below are the scenarios we discuss in the pilot; we confirm the set of functions for your site.
-
Hard hat and helmet liner
Crane, assembly and loading zones. A person in the frame without a hard hat — an event for the area foreman, with a frame and a time.
-
High-visibility vest
Forklift routes, warehouses, in-plant transport paths. A missing vest is visible on the camera from a distance.
-
Mask and safety goggles
Paint shops, food production, laboratories. Control of mask and goggle wearing where the occupational safety instruction requires it.
-
Danger zone
A person under a load, next to a running machine or behind a barrier. Entering the zone is flagged regardless of whether the person wears a hard hat.
-
Rules by time
A rule is tied to a shift: a hard hat and vest are mandatory during working hours, and at night there should be nobody in the workshop.
-
A report for the occupational safety service
Violations by zone, shift and PPE type: where to move a sign, put up a barrier or hold an unscheduled briefing.
Video analytics for retail: visitor counting and queues
-
Traffic and conversion
The number of visitors by hour and day. You can see when the shop is full but there are no sales, and when staff are short. The comparison with the number of receipts — from the data of your till system.
-
Queues at the tills
The system can count people in the till area and send an event when the queue is longer than the set limit: the manager opens another till before the customers leave.
-
Zones and time at a display
How many people came up to a promo stand and how long they spent at it. Gender and age group — in anonymised form, for the report on the shop's audience.
-
Staff areas and the stockroom
A stranger behind the counter or in the stockroom after closing, an open back door — an event for the shop security.
Customer analytics in a shop from the cameras that already hang above the entrance and the tills. People are counted, not recognised: the report holds numbers and times, without faces. We refine the metrics in the pilot.
Visitor counting does not match faces against a database and does not recognise regular customers. Recognising customers by face is a separate scenario: only against a list of those who have given consent (the “What data each function processes” block).
Video analytics for site security: zones, plates, lists
Video analytics for site security takes the routine off the guards: watching an empty perimeter and scrolling through the archive after an incident. We refine the scenarios below in the pilot.
-
Perimeter and restricted zones
A person crossed the fence line, entered a zone after closing or lingered at the server room door. The guards receive an event with a frame, not a sensor signal without a picture.
Suitable for:
for warehouses, industrial sites, car parks
-
Entry and gates
Events at the entrance: a vehicle at the gate outside working hours, movement at the checkpoint. Reading vehicle licence plates and an entry log are discussed for the project — which formats are needed and whether the product has them.
Suitable for:
for checkpoints, logistics centres, car parks
-
An alert by a list
A face in the frame matched a person from your list — the guards receive an alert and decide for themselves what to do. The lists are only those you have compiled, with a lawful basis for each person.
Suitable for:
for sites with their own security service
-
Access by face at the entrance
If a camera is installed at the checkpoint, video analytics is complemented by face-based access control: the passage opens for employees from the access list who have given consent. More on the page on Face-recognition access control.
Suitable for:
for business centres, factories, warehouses
What data each function processes
Not all video analytics is biometrics. Personal data becomes biometric when a face is used to establish who is in front of the camera (Part 1 of Art. 11 of 152‑FZ). Most scenarios do not do this — and a conversation with a lawyer is easier to start from this table.
- A hard hat, vest, mask, goggles on a person in a zone
- Entering a zone, crossing a line, movement at night
- A vehicle plate at the gate and an entry log
- The number of people at the entrance, in a queue, in a zone
- Time at a display, hall occupancy by hour
- Gender, age group, emotion — in a report without faces and without storing templates
- A camera archive: a person in the recording can be identified, although the system does not recognise them
- The retention period of the archive and events, and the circle of people with access — in a local regulation
- Cameras in a workshop for monitoring work safety — the employer's right under Art. 214.2 of the Labour Code of the Russian Federation
- Matching a face in the frame against a list of employees, guests or customers
- Photo search in the video archive
- Recognising everyone with no list and no consent — we do not do this
- “Without establishing identity” — the system sees an object or an event, not a person
- “Anonymised” — the report holds numbers, faces are not stored
- “Biometrics — consent” — the person's written consent and a lawyer's decision under 572‑FZ are needed
- “A lawyer decides” — the classification depends on the scheme and the purpose of processing
What NeuroVision video analytics does not do
-
Does not recognise everyone
A face is matched only against your lists. A person who is not on the lists remains a “face with no match” for the system — nobody establishes their name.
-
Does not search for people on the internet
There is no search across social networks and open sources, and there will not be. The source is only your cameras, your archive and your lists.
-
Does not punish for violations
An event is a reason for the foreman to come over and talk, not an automatic fine. Disciplinary decisions are taken by the employer under the Labour Code; video analytics only shows the frame.
-
Does not keep working time records by camera
We do not offer clock-in and clock-out marks from video surveillance cameras without the employee’s participation. For time tracking there is a separate scenario with the employee’s consent — time and attendance by face.
Video analytics is useful as long as employees, customers and lawyers trust it. That is why the system has boundaries that cannot be configured away.
Access to events, the archive and face images is separated by roles: the occupational safety engineer sees their zones, the guards — their cameras. How to protect biometrics in your own systems — in the article “How to ensure the protection of personal and biometric data in KYC processes”.
Video analytics and the law: the Labour Code, 152‑FZ and 572‑FZ
Three groups of rules matter for video analytics at work and in a sales area: the Labour Code on occupational safety and video recording, 152‑FZ on personal data and the restrictions of 572‑FZ for face recognition. Below is a summary of the rules as of October 2026: a guide for a conversation with a lawyer, not a legal opinion.
- The employer may use, to monitor the safety of work operations, devices and systems for remote video, audio or other recording of work processes, and store the information obtained (Art. 214.2 of the Labour Code of the Russian Federation)
- In its letter of 19.07.2024 no. 14-6/OOG-4414 the Ministry of Labour repeated this right and stated that the lawfulness of the employer's actions in a specific case is established by a court and the bodies that consider labour disputes
- The purposes, camera locations, retention period of recordings and the circle of people with access are set out in a local regulation, and employees are made familiar with it against signature
- The employer is obliged to provide employees with personal protective equipment and to organise control over its correct use (Art. 214 of the Labour Code of the Russian Federation)
- The employee is obliged to use correctly the personal and collective protective equipment issued (Art. 215 of the Labour Code of the Russian Federation)
- The procedure for issuing and using PPE — the Rules for providing employees with PPE, order of the Ministry of Labour of Russia of 29.10.2021 no. 766n
- Biometric personal data is information about a person's physiological and biological characteristics that the operator uses to establish identity (Part 1 of Art. 11 of 152‑FZ)
- Scenarios without matching a face against a database do not establish identity. How to classify a video recording and anonymised attributes in your scheme is for a lawyer to decide
- A person's image taken in places open to free access may be used without their consent if they are not the main object of use (subpara. 2 of para. 1 of Art. 152.1 of the Civil Code of the Russian Federation)
- This is a rule on the use of an image, not permission for biometrics: visitors may not be recognised by face without their consent (Part 1 of Art. 11 of 152‑FZ)
- Visitors are warned about video surveillance at the entrance, and anonymised reports are not linked to loyalty cards
- Biometrics is processed only with written consent, except for the cases in Part 2 of Art. 11 of 152‑FZ — for example, under the legislation on transport security or on countering terrorism
- Organisations are prohibited from collecting and storing in their systems biometrics used for identification, except for collection for the Unified Biometric System (item 1 of Part 1 of Art. 15 of 572‑FZ). Whether your list and search in a stream fall under the prohibition depends on the scheme — a lawyer gives the classification before launch
- 572‑FZ does not apply to matching with the participation of an authorised official (item 2 of Part 2 of Art. 1 of 572‑FZ) — the “alert, and the guard decides” mode is built with this in mind
- The operator is obliged to take measures to protect personal data (Art. 19 of 152‑FZ)
- The data retention period and the place of its processing are set in the contract; personal data can be deleted on request via the API
- Access to events, the archive and face images is separated by roles
- The summary of the rules is simplified: the Labour Code of the Russian Federation — as amended on 25.05.2026, 152‑FZ — on 26.07.2026, 572‑FZ — on 28.12.2024, the Civil Code of the Russian Federation — on 10.06.2026; check the current versions in ConsultantPlus. What the UBS is and why it matters for biometrics — a blog article.
Industries where video analytics works
Where there are more cameras than people able to watch them.
-
Manufacturing and construction
- PPE control in workshops and on site
- Danger zones near cranes and machines
- Reports for the occupational safety service Learn more
-
Retail chains
- Visitor counting and conversion
- Queues at the tills
- Staff areas after closing
-
Warehouses and logistics
- High-visibility vests on forklift routes
- Vehicles at the gate
- Movement in the warehouse at night
-
Transport and stations
- Passenger flow monitoring
- Occupancy of halls and platforms
- Events in staff areas Learn more
-
Business centres and hotels
- Occupancy of entrances and lift lobbies
- Staff premises
- Access by face for staff Learn more
-
Public sector and restricted sites
- Data storage requirements — in the contract
- Russian software from the registry
- Access to events and the archive — by role Learn more
Cameras, video server and archive: what the analytics connects to
NeuroVision provides the analytics and the API. Cameras, the video server and installation are yours or your integrator's. We choose the camera requirements for the task and check them in the pilot: they differ for overview scenarios and for face recognition.
-
Cameras
Cameras already installed, with a video stream over the network, are suitable; we check compatibility and the minimum requirements in the pilot. Face recognition needs a camera that sees the face large and frontally — the face size and the turn angle can be limited in the source settings.
-
Video surveillance system
The analytics works alongside your video surveillance system. How exactly to receive video and return events — to the account, via the API or to your system — we determine at connection; changing the video surveillance system is usually not necessary.
-
Archive and video files
Recordings from past days are uploaded as files — they are used to search for events and check how the analytics works even before live cameras are connected.
-
An integrator
Connection on site is done by your integrator or a NeuroVision partner: they know the cameras and the site network, and we provide the API and help to configure the rules. For partners and integrators
Connection: the API, alerts and your video surveillance system
We start with a pilot: we connect a few cameras or an archive, configure the rules and look at the events together. What of this is available via the API, we clarify for your project.
- Cameras and archive
- Lists and alerts
- 1 month test
- Documentation and demo
Terms of connection
A 1 month test, support under the contract, payment per check.
How much video analytics costs
Video analytics can be bought for a task: the price depends on the number of cameras, the scenarios and the connection scheme. You can start with a free pilot.
-
A few cameras or an archive, one or two scenarios, setting up zones and rules, reviewing events together with your service.
- Calculate the cost
Faces by lists
on request
Matching faces in video and in the archive against lists of people who have given consent, alerts on a match and access roles.
- Calculate a project
Video analytics for a project
on request
We determine the number of cameras, the scenarios and the connection scheme together after the pilot. The terms are tailored to your site.
Cameras, the video server and installation are not included in the price. Prices for online document and face checks are in the price matrix; it does not include video analytics — we calculate the cost for the project.
Glossary: video analytics, detection, zone and threshold
Terms you will meet in the project, in the contract with the integrator and in a conversation with the occupational safety service.
- Video analytics
- Automatic analysis of video by a neural network: finding people, objects, events and faces by rules.
- video analytics
- Intelligent video analytics
- AI video analytics that works with real-time video and with the archive, and does not just record.
- AI video analytics
- Detection
- Finding an object in the frame — a person, a hard hat, a vehicle — without establishing identity.
- object detection
- PPE
- Personal protective equipment: a hard hat, vest, goggles, mask, gloves and others.
- PPE
- Zone and line
- An area or a boundary on the camera frame for which a rule is set: entry, crossing, time spent.
- zone, line crossing
- Visitor counting
- An anonymised count of people at the entrance or in a zone, without matching faces against a database.
- people counting
- Face recognition in a video stream
- Matching faces in the frame against a list of people who have given consent. More on the page on face recognition.
- face recognition in video
- Event
- A record of a rule being triggered: type, camera, zone, time and frame.
- event, alert
- Trigger threshold
- The minimum confidence of the neural network at which an object or a match is recorded. The higher the threshold, the fewer false events but the more misses.
- detection threshold
- Source
- A camera, video stream or archive file connected to the analytics.
- video source
Licences and certificates
The project «Development of a multimodal system using artificial intelligence "Neuromed" for managing the diagnostic processes of a clinic on the basis of the "Neurovision" computer vision platform» is supported by the Innovation Promotion Foundation within the federal project «Domestic Solutions»
The Neurovision computer vision platform is registered with Rospatent (certificate No. 2023661508) and included in the Russian Software Registry (entry No. 19972 of 13.11.2023)
The right to use the Platform is granted under a licence agreement concluded by Neurovision with its customers
We will choose video analytics for your cameras
Tell us what the tasks are and how many cameras, whether you have a video surveillance system and an archive — we will propose scenarios and a pilot on your recordings.
Can't put your request into words? Book a call and our team will contact you at a convenient time
Book a call
The company is included in the register of IT companies and the register of personal data operators of the Russian Ministry of Digital Development.
Address: Moscow, Presnenskaya Embankment 12 — Moscow City business centre, Federation Tower. Nearest metro stations: Delovoy Tsentr and Mezhdunarodnaya.
Read also
Materials on face recognition, algorithm accuracy and protection against deepfakes — in our blog.
156 23
KYC, AML and Digital Identification
Face Recognition System: How It Works, Its Accuracy, and Where It’s Used
380 23
KYC, AML and Digital Identification
Local or cloud KYC: how to choose a biometric data storage model
384 16
KYC, AML and Digital Identification
How to ensure the protection of personal and biometric data in KYC processes
167 18
KYC, AML and Digital Identification
Unified Biometric System (EBS): What It Is, How to Enroll Your Biometrics, and Where It’s Used
470 28
KYC, AML and Digital Identification
Face recognition accuracy in KYC: comparing biometric algorithms and face recognition errors
518 29
KYC, AML and Digital Identification
How Video Streams Are Spoofed in KYC: The Architecture of Attacks via Emulators, Virtual Cameras, and Video Stream Injection
Frequently asked questions about video analytics
Video surveillance records video, video analytics analyses it. A neural network works with the video from cameras and the archive and flags events by your rules, and matches faces against lists. The operator does not need to watch all the screens — they receive an alert and a frame.
Most often not, but it depends on the task. We work with the video from your cameras and with archive recordings; we check compatibility and requirements in the pilot. Face recognition needs a camera that sees the face large — we choose it for the location.
Zones where a hard hat, vest or goggles are mandatory are marked on the frame, and a rule is set. If the violation lasts for the set time, the responsible person receives the event with a frame and a time. The system does not determine who broke the rule. We clarify the capabilities for each type of PPE in the pilot.
No. Visitor counting by video analytics is anonymised: the report holds numbers, queues and time in zones, gender and age group — without faces and without matching against a database.
This is a scenario we discuss for the project: write which entrances and plate formats you need — we will answer what is available now and what can be checked in the pilot. How long to keep the entry log is determined by your local regulation.
Yes, but only against your lists of people who have given written consent — employees, guests, loyalty programme customers. The source can be a camera, a photo or an archive video. The system cannot recognise everyone and will not; your lawyer checks a scheme with face recognition before launch.
Yes, this is a separate scenario — access control with face recognition: a camera at the checkpoint, access lists, consent and a card for those who have refused biometrics. More on the page face-recognition access control.
To monitor the safety of work operations the employer may use remote video recording (Art. 214.2 of the Labour Code of the Russian Federation); the purposes and procedure are set out in a local regulation, and employees are made familiar with it against signature. If faces are matched against a list, however, that is biometrics and written consent is needed (Art. 11 of 152‑FZ).
We do not offer time tracking by video surveillance cameras without the employee's participation. There is a separate scenario for that — time and attendance by face at a terminal, with the employee's consent.
The first month of a pilot on your cameras or archive is free. After that we calculate the cost for the project: by the number of cameras and scenarios. Cameras, the video server and installation are separate, with you or your integrator.
