Player Location Estimation for Occluded Broadcast Video Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional player tracking techniques are limited to in-venue tracking and fail to accurately estimate the locations of players not visible in video broadcasts of sporting events, which are essential for calculating statistical measurements and analyzing game strategy.
Innovation Solution
A computing system uses a prediction model, trained with in-venue tracking data and broadcast video data, to estimate the locations of players not visible in video feeds by employing spline models and nearest neighbor approaches, such as kNN, to interpolate or predict missing player locations based on known positions at the start, end, or during possessions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If camera zooms into where the main action is, then visibility of players closer to action is improved, but visibility of players at strategic locations is lost
Solution Approach 1:
The patent uses prediction models as an intermediary to bridge the gap between visible and invisible player locations. The model takes visible player positions as input and generates estimated positions for players not currently in view, effectively mediating the information loss caused by camera zooming.
Solution Approach 2:
The system performs preliminary tracking and recording of player positions during periods when players are visible, storing this data for later use. This preliminary action enables the prediction model to estimate positions during periods when players go out of view, preparing information in advance to compensate for upcoming visibility gaps.
2Device complexity
If conventional player tracking is used in video feeds, then implementation simplicity is maintained, but accuracy of player location estimation for invisible players deteriorates
Solution Approach 1:
The patent replaces conventional mechanical/optical tracking systems with a computational prediction model. Instead of relying solely on physical camera coverage and optical detection, the system uses machine learning algorithms to predict player positions, substituting computational methods for traditional mechanical tracking approaches.
Solution Approach 2:
The system creates a virtual copy of the physical playing environment using prediction models. This digital replica estimates player positions based on available data, allowing analysts to work with a complete virtual representation of player locations even when physical cameras cannot capture all players simultaneously.
Data Source
AI summary
Examples disclosed herein may estimate locations of players not visible in a sporting broadcast video. A prediction model may be generated based on a training data set of in-venue tracking data that includes locations of all players at all times and the corresponding broadcast tracking data that may not necessarily contain the locations of all players at all times. The prediction model may be based on an algorithmic logic (e.g., a spline regression) or machine learning model (e.g., k-nearest neighbor, deep neural network). The generated predicted model may be used to estimate the unknown locations of players in broadcast tracking based on the known locations.