Computer Vision (CV) is a subdivision of machine learning that develops the techniques to train computers to understand the content of images.
Since a video is also a combination of pictures, the same methods can also be applied. CV tries to replicate the complexities of the human vision and couples using deep learning models to detect and classify objects from the dynamic world.
Need for Automation in Sports
By their very nature, sports comprise fast motion that is difficult for competitors to master, trainers to decipher, and spectators to follow. It is impossible to track and monitor the players by attaching sensors or fixed devices with them.
The use of Computer Vision techniques has filled this lacuna.
Systems based on Computer Vision are being used in the commercial sports industry.
Current Application of Computer Vision in Sports
Less than five years ago, the sports industry was utterly alien to the concept of Computer Vision.
Today, however, deep learning and computer vision techniques are being used by broadcasters to enhance the experience of the audience and even by the clubs themselves to gain a competitive edge over their opponents.
Computer Vision systems are correctly able to distinguish between the ground, players, and other objects. Algorithms make the use of color-based elimination to separate the background and foreground. For example, color-based segmentation detects grass due to its green color and declares it as the scene's background with players and objects moving in front.
Player Tracking and Field Placement
One of the most widely used applications of CV in sports is player tracking. The player tracking enables coaches to improve the performance of their teams and determine the ideal on-field formation. The results obtained from a CV
system provide detailed and robust statistical analysis of the individual players and team performance. Automated suggestions are then provided to the coaches taking into account the opposition.
They can also compare the actual positioning with the suggested ones. Event recognition techniques detect events such as goals, penalties, shots, and near misses using a contextual environment such as court color or lines on the pitch. Predefined themes can also be set to generate specific content for easy post-match browsing.
Automated Ball Tracking
Racket and bat-and-ball sports such as squash, tennis, badminton, or cricket have been using computer vision since the mid-2000s. Ball tracking systems index through images to identify all objects that resemble the elliptical shapes of a ball. Upon detection, a three-dimensional trajectory is constructed by linking multiple frames to project the ball path. This system determines whether the ball fell within the boundary or not. Tennis games make the use of this technology to determine the legitimacy of the point. In cricket, the predicted path of the ball is displayed had the batsman not hit it.
Commercial Usage
IBM developed a system for Wimbledon in 2017 that was able to generate automated match highlights. It took into account crowd noise, player movement, and match data to determine the entire match's pivotal moments. FIFA uses a seven camera Hawk-Eye system based upon Computer Vision. It uses high-speed cameras and a goal detection system that covers the entire goal area. This extremely accurate system allows referees to determine whether the ball crossed the goal line or not.
Current Challenges
Despite the widespread use and potential of Computer Vision in sports, some critical problems need to be resolved before the full benefit can be reaped. Varying body postures, occlusion of players by objects are some of the pain points CV has not yet fully addressed. The sudden unexpected motion of players, the similarity between players, and close interaction pose challenges for the tracking system.
However, despite the challenges, Artificial Intelligence continues its growth due to the massive amount of money poured by tech giants such as Google, Amazon, and Intel. With the passage of time, this technology will eventually mature, and the dream of full automation will be achieved.
