TU/e Visualisation Project – High Volume Sensor data

This paper describes the findings and development process of the TU/e Visual Computation course. This paper presents the implementation of a visualization-system that is able to use various datasets of XYZ-positional body-sensors and corresponding videos from several professional soccergames. The dataset includes anonymized player-data among which are time-stamped field positions, speed, acceleration and directional vectors. The datasets can be used to provide insights in player’s performance and spectator statistics. To accomplish this, computer-vision algorithms that implement techniques such as feature extraction, object-tracking, background subtraction a.s. can be utilized alongside the video-feed which enables the visualizations to be tested against a ground truth. Furthermore an evaluation, discussion of results and future expectations were given in this paper.

Description

  • Eindhoven University of Technology

  • Paper