Ball-Holding Player Tracking via Pose-Ball Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional optical detection techniques are inadequate for accurately tracking a ball held by a player in live sports games, especially when the ball is occluded or due to its small size and high speed, and mounting RFID sensors can alter the ball's kinetics and increase costs.
Innovation Solution
A computing system that automatically identifies and tracks a ball-holding player by analyzing two-dimensional video data from multiple cameras, using techniques such as YOLO for ball detection and deep-sort for player tracking, and a pose-ball model to estimate the ball's location based on the player's body parts, eliminating the need for RFID sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional optical detection techniques are used to track the ball, then the system is simple and non-invasive, but the tracking accuracy deteriorates when the ball is occluded or due to its small size and high speed
Solution Approach 1:
The patent introduces player tracking as an intermediary method to infer ball position. When the ball is occluded or difficult to detect directly, the system uses the tracked position and motion characteristics of the player holding the ball to estimate the ball's location, thereby maintaining tracking accuracy without adding complex sensors to the ball itself
Solution Approach 2:
The patent replaces direct optical detection of the ball with a computational approach that substitutes the detection problem with a tracking and inference problem. Instead of trying to optically detect the small, fast, occluded ball, the system uses player tracking data and ball physics models to substitute for direct ball detection, achieving better accuracy
2Measurement precision
If RFID sensors are mounted on the ball to improve tracking accuracy, then ball tracking accuracy is improved, but the ball's kinetics are altered and costs increase
Solution Approach 1:
The patent extracts the tracking function from the ball itself and relocates it to the player tracking system. Instead of equipping the ball with RFID sensors that would alter its kinetics, the system extracts ball position information indirectly through player tracking and spatial reasoning, eliminating the harmful effect of altering ball physics
Solution Approach 2:
The patent creates a virtual copy of the ball's position and motion trajectory by inferring it from player data and ball physics models. This virtual representation allows the system to track the ball without physical contact or sensors on the actual ball, preserving the ball's natural kinetics while achieving accurate tracking
3Measurement precision
If RFID sensors are mounted on the ball to improve tracking accuracy, then ball tracking accuracy is improved, but the system costs increase
Solution Approach 1:
The patent uses inexpensive player tracking data and open-source computer vision algorithms instead of expensive RFID sensors. The system leverages existing camera infrastructure and processes player position data through computational methods, dramatically reducing system cost while maintaining or improving tracking accuracy compared to RFID solutions
Data Source
AI summary
Methods, systems and apparatuses may provide for technology that selects a player from a plurality of players based on an automated analysis of two-dimensional (2D) video data associated with a plurality of cameras, wherein the selected player is nearest to a projectile depicted in the 2D video data. The technology may also track a location of the selected player over a subsequent plurality of frames in the 2D video data and estimate a location of the projectile based on the location of the selected player over the subsequent plurality of frames.