August 26, 2024
Computer vision and artificial intelligence in manufacturing
A few examples of how modern technology makes it possible to produce things faster, cheaper — and more environmentally friendly
.jpg)
First of all, computer vision technology helps to recognize players accurately. Not only visual recognition is used (the AI determines the color of the uniform, number, and name on the jersey, as well as the face), but also a combined approach — camera data is supplemented with a priori information, such as the position and roles of players on the football field.
Most modern stadiums (and all commercial European clubs) already have Hawk-Eye systems in the stadiums — making player recognition and tracking continuous and accurate. 14 AI soccer cameras are enough to cover the entire football field.
Today, Hawk-Eye is the leading computer vision technology in soccer and sports. It is used in more than 500 stadiums in 90 countries, and artificial intelligence at football games is used for more than 20,000 matches in 90 countries around the world every year.
Such a complex system can track minuscule movements and help with advanced analysis. However, the widespread availability of high-resolution cameras (up to 4K and beyond), that can be positioned above the stadium, allows for confident player recognition even with a single camera setup. Many clubs can afford such a system, as the cost of the camera and software does not exceed several thousand dollars.
By recognizing every player on the field and tracking them in real time, the AI helps gather a rich profile for each player. Here's what you can learn at every moment of the match for example:
• The player's running speed and acceleration: peak and average.
• The distance he ran during the match.
• The reaction time to a pass or an open attacking moment.
• The zone of visibility and the players who are available for a pass.
• The pattern of movement of the football player on the field.
• The position of the player’s hands and feet while striking the ball. This allows us to assess the quality of the kick (there are several types of kicks in football).
This data is also available to viewers to enrich broadcasts with visualizations. And also for coaches — right during the game.The data can be transferred to special software for trainers, such as Catapult.
The large amount of data that AI provides can be used in unusual ways. For example, in the sports market, some companies are professionally engaged in predicting sports matches — using a large amount of data not only about the team but also about the players in particular. Using a probabilistic method, Kickoff.ai achieves a prediction accuracy of 97.9% (though only for data-rich matches). Thus, artificial intelligence predictions in football are not the future, but the very present.
AI-assistants also help coaches to create an optimal team composition and suggest tactics that are best suited for a match against a particular opponent. For example, AI is helping coaches at Liverpool Football Club select the right players for corner kicks - based on analyzing the performance of more than 7,000 corners taken in the team's matches.
Player data from computer vision and artificial intelligence pumps up training and helps each athlete reach peak performance.
Developing physical capabilities is a real-world task where AI is not needed. Another thing is the spatial thinking of football players, their ability to anticipate the situation on the field.
Good football players are like chess players, they keep in mind the potential for a breakthrough, kick, or pass, analyzing them on the fly and choosing the best one. Computer vision can help with this. How a computer builds a field of opportunity for football players.
For example, by analyzing players and their positions on the field, AI can make assumptions, suggesting the best passes and kicks. Players and coaches revisit visualizations of matches, clearly identifying missed opportunities.
Computer vision improves player performance — it is a fact. For example, Liverpool players became 5% faster after coaches started analyzing data from training sessions and matches.
This helps to “work out” synergy that didn’t happen on the field — in the next match, player A will pass to player B more actively because the AI has seen that this guy understands and is more likely to be ready to pass.
Computer vision and machine algorithms are already integrated into top-tier leagues. Today, AI-assisted referees are used everywhere in FIFA matches (especially since all stadiums are already equipped with Hawk-Eye technology).
Human error in detecting offsides can be up to a meter — it is very difficult to estimate the actual position of several moving players relative to a parallel goal line. At the same time, the accuracy of Foxtenn's Hawk-Eye system is a few millimeters. The millimeter accuracy of the position of the ball is impressive. Such accuracy is achieved by shooting at 2500 frames per second!
AI is coming into refereeing primarily because of the high cost of error. For example, up to 27% of football match decisions made by referees were based on misconceptions. Most often, referees are wrong about offsides and assessing unsportsmanlike behavior situations.
Auto-offside detection technology worked out at the 2022 FIFA World Cup in Qatar. However, in addition to computer vision, they used a special Rihla Pro ball with sensors installed in it. However, CV already makes achieving no less accurate ball positioning possible.
Recognizing the ball on the field is quite a challenge from a computer vision perspective. The ball is small, moves very fast, and can have many kinds of patterns. To recognize a ball without errors, complex mathematical problems have to be solved.
The augmented reality technology for referees is called VAR — Video Assistant Referee. The referees on the field can refer to VAR in case of doubt. For example, during an important match in the German Bundesliga, VAR was called 12 times, and in one case the referee decided on a fateful penalty based on a hint from the system.
Modern AI systems in refereeing can analyze not only the position of players on the field but also accurately determine the position of players' hands and feet. For this purpose, 3D models of the players are built based on camera images, also with millimeter accuracy. Such technologies are already widely used in gymnastics, where artificial intelligence actively judges athletes. Their widespread use in football is a matter of a few years.
The most important thing about football is its spectacle. And here, computer vision technologies are also working at full capacity — with soccer artificial intelligence.
During any modern broadcast, several “layers” of additional data appear. Not only usual replays — the broadcast can change the angle of view dynamically, thanks to a 3D scene reconstructed from cameras.
The broadcast is enriched with infographics around the players, showing the speed and direction of the ball kick. We can pause the broadcast, rotate the field to view the players, and then accelerate to the real-time moment. The magic of technology!
We at OpenCV.ai specialize in computer vision technologies — in sports too. Read more about our solutions. And let us get to know your challenges — we'll use our cutting-edge technology to help you achieve your computer vision and artificial intelligence goals
.jpg)