Using Video Analytics to Record HSE Violations and More
Despite the downward trend in industrial injuries, their level remains quite high. A significant portion of accidents occur as a result of employees violating HSE requirements, as well as due to personal negligence.
In their pursuit of reducing industrial injuries, employers make significant efforts and spend substantial funds on technical measures. However, none of us are immune to accidents caused by the personal negligence of employees or intentional disregard for HSE requirements.
Given the successful experience of using video analytics for commercial purposes and ensuring security at various facilities, our company decided to implement a video analytics system to record instances of employees entering danger zones or areas with operating equipment.
Video analytics is currently used not only in "Smart City" projects and warehouse sites in static spaces but also at production facilities with active equipment. In the case of production facilities, the main technical challenge is that moving parts of equipment fall into the cameras' field of view, which the system might mistake for a person. To solve this problem, the client, together with the developer, creates a so-called "dataset" — a database used for further training of neural networks.
One version of the current system for detecting the presence of employees in danger zones is presented below.
The technical solution consists of two main components:
- Cameras with autofocus support installed in danger zones.
- A smart recorder with video analytics capabilities installed in the server room.
Action scenario:
- The camera captures an image, which is simultaneously transmitted to general-purpose video surveillance servers and the smart recorder.
- The smart recorder records only the moments when people are in danger zones and sends an email notification to the relevant parties.
- The message contains the name of the workshop and the section where the violation was recorded, as well as a link to the video file.
Equipment built on neural networks does not stop at the results achieved. The system can be trained to perform various tasks without changing the hardware, but by upgrading the existing setup with modules developed for specific tasks.
The range of tasks that can be solved using video analytics is diverse — from monitoring the use of personal protective equipment (PPE) to fire detection.
