Set Up
Full Rink Coverage
Our system will use a series of 6 cameras that are placed above the ice on rotating arms. Four will be placed on each side of the rink, while two will be behind each net. They all are connected to and controlled by a central computer in the control room, where an employee inputs arena dimensions and constraints. The purpose of these cameras is to watch the game and track data points and movement across the ice (players, puck, and arena layout), in order to precisely call offsides. If the cameras detect a player crossing over the blue line into the offensive zone before the puck does, a horn will sound in the arena, effectively stopping play. But to do, the technologies we employ include:
Artificial Intelligence (AI): Computer systems able to perform tasks that normally require human intelligence, in our case visual perception and decision-making.
Machine Learning: A subset of AI, allows our computer algorithms to automatically learn and improve without manual programming.
Vision Sensor: Camera processor that uses position, orientation, contrast, and lightning to determine and track motion.
Functionality
State of the Art Vision Sensors with Artificial Intelligence Processing in the Sony IMX501 chipset
At the heart of our cameras, is the brand new Sony IMX501 intelligent vision sensors, the first image sensors in the world to be equipped with AI processing functionality. The issue with vision sensors now is that they process every piece of data in the frame, which takes too much time for the situation were approaching. High-speed edge AI processing, extracts only the necessary data which would be the players, puck, and arena layout, in turn, reducing data transmission latency and power consumption. Every millisecond, an image of these data points is captured. Pictures from each camera are compared against each other to determine if a player is in the offensive zone before the puck.
Application
Machine Learning in Charge of Purposeful Data
Being able to differentiate data is one thing, but understanding the data to make the offsides call is our goal. To do this we use a subset of AI, defined Machine Learning. Repetition of captured frames from in-game situations and data algorithms allows the AI and main computer to adapt and become smarter over time. First, we need a baseline, which is built by feeding the computer hundreds of thousands of theoretical in-game simulations that depict offsides. Over time, our product will become smarter and even faster in detecting offsides and essentially be able to run itself. This makes for an intuitive and cost-efficient user experience, as the only inputs include game-specific data points like team color. The storing and gathering of data will not just prove the accuracy of our product, but add value, in turn, generating more sales.