VisionLabs connects Luna Cars to MTS office parking: camera-based access control

VisionLabs has deployed a parking automation system at the MTS office built on its Luna Cars video analytics. According to CNews, the system controls vehicle access: cameras capture the car, computer-vision algorithms identify it, and the system then decides whether to open the barrier. The publication, dated 13 October 2026, gives no number of parking spaces, no recognition accuracy, and no deployment timeline, so what can be assessed is the format of the project, not its results.

What is known about the project

The customer is MTS, one of Russia's largest telecom operators. The contractor is VisionLabs, a Russian developer of computer-vision systems. It has long worked in face and object recognition, and Luna Cars is part of its product line for vehicle-related tasks.

A corporate parking lot is a natural use case for this technology. Traffic is predictable, the list of authorized vehicles is limited, and cameras can be placed so that the angle and lighting stay stable. These are the easiest conditions for video analytics: an error here costs a single delay at the barrier, not an incident on a city road.

How systems of this type work

The typical setup looks like this. A camera at the entrance captures a frame or a video stream, a model finds the car and its license plate, and then it recognizes the characters. The result is checked against a list of permitted vehicles, and the system sends a command to the barrier. Solutions of this type can often also detect additional attributes: make, color, body type. The source does not say which functions are enabled at the MTS office.

The main advantage of this approach is that it needs no physical media. There are no passes or key fobs, and no attendant checking plates by hand. Guest visits can be arranged in advance: the plate number is entered into the system before the visitor arrives.

What the report leaves out

Basic data needed to judge the project's maturity is missing:

• Scale: the number of parking spaces and entry lanes is not stated.
• Accuracy: the share of correctly recognized plates in different conditions (night, snow, dirty plates) is not given.
• Speed: the time from a car appearing in frame to the barrier opening is not stated.
• Integration: it is unclear whether the system is linked to the corporate access control system and employee lists.
• Economics: the cost of deployment and the payback period are not disclosed.
• Data: the report does not say where camera recordings are stored or for how long.

Without these figures, the report remains a showcase case. For VisionLabs, such a case is valuable for its reputation: a high-profile customer and a clear task. For the market, it does not yet provide material for comparison with competitors.

What to check before deployment

If a company with its own parking lot plans a similar system, it should start with a pilot on a single lane. Measure the error rate over a month, including bad weather and winter months, when plates get dirty and low sun blinds the cameras.

Check failure scenarios separately. What happens when the connection to the server is lost? Can the barrier be opened manually? How are unfamiliar and foreign plates handled? Third, personal data: a license plate is often tied to a specific employee, so you need a retention policy and restricted access to the archive.

If the vendor is ready to state accuracy and latency in numbers and put them in the contract, that is a good sign. If it stops at the word "automation," ask for test reports.