Like the Video Assistant Referee, the Maruzsi System is based on image analysis, but that is the only similarity. All the elements of the Maruzsi System are specifically developed for football stadiums, customised to the accurate dimensions and realistic 3D surface of football pitches. It automatically detects when the ball crosses the goal, base, or sidelines and automatically recognises offsides. It sends signals to the linesman’s newly developed flagpole, which flashes its LED lights, beeps, vibrates accordingly, and even shows the offside graphic on its display. The processing time of VAR may take minutes; the Maruzsi System provides the results immediately with real-time analysis. Decision making support is given back to the linesmen, so the game may continue without lengthy interruptions, just as before VAR. The accuracy of the system is one centimetre.
The 3D surface of the football field
The first step of implementing the Maruzsi System is to measure the exact dimensions of the stadium and the football field, the slope angles of the pitch, and possible distortions. We create a 3D model of the football field based on these data, so the software can calculate with the actual, curved spatial surface instead of a theoretical flat area. This is of great importance in the light of this:
Football pitches are not horizontal flat surfaces for rainwater drainage reasons but humps or mounds sloping downwards from the longitudinal axis. The corners are 30-40 cm lower than the axis. This slight slope is not visible to the naked eye and does not affect the ball or the players, but it does affect the evaluation of offsides.
When VAR referees analyse a possible offside situation and draw the imaginary straight VAR line on the camera images, they assume the football pitch is a horizontal, flat surface. If this were true, the actual offside line and the VAR line would overlap, showing the situation correctly. However, the physical football field has a slightly curved, 3D surface, which the actual offside line would follow if it were physically there. But it isn’t, so VAR referees do not notice that their straight lines are not measuring reality but a slightly distorted version of it. They fit the straight line to the defender and falsely call an offside. See the illustration in this 14-second video. (For clarity, we mildly scaled up the football pitch along the vertical axis.)
In reality, the deviation can be 10-20 centimetres, depending on where the offside is, the angle between the camera and the offside line, the exact slope angle, and the field’s dimensions. The VAR image gives accurate results only if the camera is precisely in line with the offside.
Maruzsi System performs the analysis based on the actual 3D surface and 3D coordinates, so its accuracy is one centimetre.
Unlike VAR, Maruzsi System does not work with a few randomly placed TV cameras. As functional accessories of the stadium, 24 to 40 high-speed, closed-circuit, fixed focal length, fixed-position cameras with at least 4K resolution are installed at precisely calculated locations in the stadium. They are mounted in pairs, close together at each selected point. The pairs are aimed at the same area but configured for different lighting conditions. This helps the system evaluate moments when one half of the ball or player is in a sunny spot, and the other half is in a shaded area.
Each camera optic is calibrated, the distortion and chromatic aberration data are recorded to compensate for lens distortion with software. This improves the accuracy significantly.
Wide-angle camera lenses are distorted by the laws of physics. This “fisheye” side effect is typical for wide-angle landscapes and cityscapes. Images like these are slightly bent around the corners; the extent of the distortion can be as great as 1-2%. An image picturing a 70 m wide football pitch with 1-2% distortion in the corners can cause a half meter deviation from the virtual straight VAR line. The Maruzsi System automatically corrects this distortion in real-time.
illustration: lens distortion
Each camera pair is assigned to a 2D image processing computer, which receives the images via a high-speed network. The 2D PC-s can analyse hundred images per second, each with eight megapixels, in 2D. The ball and player data are then sent to the 3D image processing server, which calculates the exact coordinates of the event and sends the signal to the linesman’s flagpole and the LED stripes on the edges of the base- or sidelines, if necessary.
The signature item of the Maruzsi System is a uniquely designed new flagpole for the linesman. It receives and displays signals from the system’s computers and informs the linesman about what happened on the field. It literally puts decision-making back in the linesmen’s hands. Built-in LED lights, sound and vibration indicators, a 70×124 millimetre display and a push-button make it unique.