◎Intern reporter Du Peng
Recently, FIFA announced on its official website that this winter's FIFA World Cup in Qatar will officially use semi-automatic offside penalty technology for the first time. The occurrence of an offside misjudgment.
In previous football games, one of the biggest concerns was that the referee misjudged offside, because even a slight misjudgment could affect the outcome of the entire game.
But now, football referees will usher in a powerful assistant - artificial intelligence.
Recently, FIFA announced on its official website that this winter's FIFA World Cup in Qatar will officially use semi-automatic offside penalty technology for the first time. The occurrence of an offside misjudgment.
It is difficult for humans to accurately determine offside with only eyes
The offside rule is not complicated. Aside from some special circumstances, generally speaking, it requires the offensive player to pass the last kick and the effective body part of the last offensive player to receive the ball must not exceed the defending team. The body of the player who is second to last from their own goal.
Although it is only a few words, but in the actual game penalty, it is undoubtedly difficult to accurately judge offside.
According to Ai Kun, a teacher at the Chinese Football Academy of Beijing Sports University and an international football referee, in order to make an accurate offside penalty, the referee must clarify the passing and receiving players at the moment of the last pass. The positional relationship between the body of the three defenders.
"Players are constantly moving all the time, and it is very challenging for the human eye to locate the positional relationship between multiple players at the moment of passing the ball." Ai Kun said, so in order to be able to more To clearly determine if a player is offside, there will be an assistant referee on each sideline in football matches, so that he can use the best viewing angle to make accurate offside calls in an instant.
However, with the continuous improvement of the level of football, the rhythm of the game has been accelerated, and the attack methods have become more and more diverse. Even if the assistant referee has an excellent observation position and eagle-like sharp eyes, it is always completely impossible to rely solely on human eyes. Accurate offside penalty has become more and more difficult.
With the rapid development of video broadcast technology, especially the wide application of high-speed cameras, the broadcast images that can be played back repeatedly not only improve the fairness of the game, but also provide the possibility to use video technology to assist in making decisions. VAR) came into being.
Ai Kun introduced to reporters that the organizer uses 10-30 cameras in the stadium to record the game in real time. When there is a key node affecting the game, VAR will provide the referee on the court by replaying the video screen. Penalty reference.
As for the offside penalty, Ai Kun said that the referee must first use VAR to "manage", that is, through frame-by-frame playback, determine the time when the offensive player sends the last ball.
Then, based on the time determined by the RBI, "line drawing" is performed to determine the relative position between the offensive player receiving the ball and the defending player who is second to last from their own goal, and draw 2D and 3D offside lines.
Although VAR has a more accurate way of observation than the human eye, its often minutes-long penalty time and nitpicky penalty scale are still repeatedly criticized by fans.
In Ai Kun's view, "Although VAR is not perfect at present, it does provide a real and objective factual basis for penalties including offside."
AI helps referees achieve "one-click penalty" for offside
In the face of VAR's unsatisfactory performance, FIFA decided to go a step further and introduce artificial intelligence technology to assist offside penalties.
Compared with VAR, which still relies on referees to manually make points and draw lines, the semi-automatic offside penalty technology that will be applied in the Qatar World Cup will completely hand over this process to the artificial intelligence system to help referees achieve "one-click penalty" for offside. .
But in the face of the offside rule that is not easy for senior fans to explain clearly, can artificial intelligence really understand it?
In fact, for artificial intelligence, it is not important whether it really understands what offside is. What it needs to do is to compare and judge the various types of data collected, and then come to an objective result.
The operating path of its system can be split into two key steps.
First, determine the timing of the offensive player's last pass.
This aspect mainly relies on Al Rihla, the official match ball of this World Cup in Qatar.
Its built-in inertial measurement unit (IMU) sensor is placed in the exact center of the inside of the ball with the support of several soft brackets, and can send the ball's running data to the decision-making center at a frequency of 500 times per second.
Johannes Holzmiller, director of technology and innovation of FIFA, said that the sensor can send the acceleration information of the ball running in multiple directions in real time. Precise time.
In the most critical part of drawing the offside line, the system mainly relies on 12 dedicated tracking cameras arranged under the stadium roof.
They will use video motion capture technology to track 29 relevant parts of each player's body, form 29 data points, and send data including location, time and other information to the decision-making center at a frequency of 50 times per second.
Finally, the system will integrate and calculate the time information returned by the ball and the player's limb position information tracked by the camera to obtain the final result.
Although it sounds complicated, the whole process only takes 3 to 4 seconds, and the referee can get an accurate result calculated by the system.
If the referee finally decides to accept the result, the system will automatically generate a 3D animation based on the data collected from 29 data points on the athlete's body, with the help of 3D human posture reconstruction technology, which will draw the offside in detail with the best perspective. The position of the line and the position of various parts of the body of the offside player when receiving the ball are shown on the giant screen in the stadium and on the broadcast screen, so that fans can be "convinced".
Since artificial intelligence is so powerful, can it directly replace the referee to make the final offside penalty?
Pierluigi Collina, chairman of the FIFA Referee Committee, repeatedly emphasized that the reason why the program is named "semi-automatic offside penalty technique" is that it can only provide the referee with the reference result of whether it is offside, but cannot make a decision. Final penalty.
Ai Kun also said that when making an offside penalty, the assistant referee must not only make a penalty based on objective facts, but in some special cases, he also needs to make a subjective judgment on the facts of the game according to the rules. Although the offensive player did not catch the ball, whether he interfered with the opponent player, participated in the attack, or interfered with the opponent player's handling of the ball by moving, etc. These penalties involve in-depth understanding of the game, and the machine cannot make a completely accurate judgment for the time being. Judge." Ai Kun said.
AI technology has been applied in multiple sports
Although the offside penalty technology has not yet been popularized in all football games, the artificial intelligence system using similar motion capture technology and 3D human posture reconstruction technology has been widely used in many sports events and training.
At the Tokyo Olympics in 2021, the Chinese diving "Dream Team" performed perfectly, winning 7 golds and 5 silvers. The AI coach is one of the heroes behind the scenes.
The coaching system, jointly launched by the Chinese National Diving Team and Baidu, is the first "3D+AI" diving training system in China.
According to Dr. Lu Feixiang, a senior Baidu R&D engineer who participated in the project, the system uses high-speed cameras to capture 2D high-definition images of athletes from stepping on the springboard to fully entering the water, and then uses brain 3D vision technology and deep neural network technology to estimate the athletes' health 3D pose, and get the 3D angle of each joint.
Then, through the 3D reconstruction technology of the human body, the whole process of diving can be reproduced in 3D.
Not just auxiliary training, the application of artificial intelligence in sports events has been increasing.
In the Beijing Winter Olympics Test Tournament held in February 2021, the artificial intelligence referee and coaching system "Guanjun" developed by XiaoIce Company served as the only competition referee in the aerial skills project.
Successfully completed the individual pre-finals, super-finals, and team pre-finals with a total of 44 referees, which were unanimously recognized by the International Snow Federation and the Winter Sports Management Center of the General Administration of Sports of the People's Republic of China, and this was the first time in human history without manual intervention. , the artificial intelligence will independently complete the task of refereeing the competition.
According to Li Di, CEO of XiaoIce Company, "Guanjun" solved the problem of athletes' action and posture recognition, and built an ice and snow sports analysis model under the condition of very scarce training data, which can analyze the movement trajectory, body posture, Introduce multi-dimensional indicators such as angle and height for analysis, and learn and simulate the scoring standards of referees in international competitions.
Regarding the future application of this system, Li Di said that the technologies shown at present are often the beginning of stages, and he believes that the application of artificial intelligence in sports training and other aspects will soon spread to ordinary people's homes.