Farseer Smart Surveillance system


The Client

The biggest power plant in Eastern Europe

The Problem

Power plant requires employees to follow very strict safety rules. The problem for this client was that the number of employees and the quantity of big facilities was sagnificant and it was really hard to watch employees follow the safety regulations. The client wanted a solution that enables automated surveillance of safety regulations, which included wearing safety glasses, safety hard hats and prevent smoking on premises.

The Solution

After negotiating with client the company came with the plan to create a module micro-services based system to analyze input data flows from different sensors based on computer vision object detetion neural network.

The tasks

  • Create a modular system for analyzing streams of video content from surveillance cameras at the facility, which will be able to carry out 24/7 control of safety rules on the basis of neural networks.
  • Use the existing network of digital surveillance cameras with image quality not lower than Full HD.
  • Connect to the video surveillance network through existing NVRs.
  • Processing capacity can be provided as dedicated servers or the cloud (Microsoft Azure).
  • The basic recognition reliability for event generation is 3 times.
  • End user operating system MS Windows 7, 10.
  • Storage of all recorded violations in a searchable database. Formation of analytical reporting on detected violations.
  • Notification of the responsible person about the event via e-mail (promotional Android application). Information in the message - time, place, type of violation, photo of the violation.


  1. Observation:
    1. List of cameras and camera groups available to the system.
    2. Display of icons of available cameras in the display panel with the possibility of setting the update frequency (stream - once every 10 minutes).
    3. The ability to watch an enlarged stream of a specific camera.
    4. Ability to create restricted areas.
    5. List of objects recognized by a specific camera \ group of cameras.
    6. A short list of recent recorded events.
  2. Reporting:
    1. A complete list of recorded events.
    2. Ability to apply filters and sorting to the list of events.
    3. Ability to make visualizations of sorted data.
    4. Ability to save/send sorted data in .xlsx, .pdf format.
  3. Plan:
    1. The plan of the room with the output of the coverage by surveillance cameras.
    2. Existing restricted areas from all cameras.
    3. Visualization of HeatMap and places of the most frequent violations.
  4. Settings:
    1. Agree as the system develops.
    2. User management.
    3. Manage the list of violation notifications.

Objects for recognition:

  • A person
  • A helmet, a man in a helmet
  • Safety glasses, a person in safety glasses
  • Open flame, smoke (how to distinguish from steam?)

As a result an automated reporting system was created based on analytics of sensor data