Look Who it Was Random video

Computer Vision Bowling Overlay

1,252 views · Published 18 July 2014 · 0:42 · Indexed 29 September 2026

Channel: vegasbryan · 2014 · Science & Technology

Watch on YouTube

The first thing happening in this video is the creation of a background matte / difference matte.  This is used to separate the player from the background.   After that, the application runs a Hough transform to find the lines in the video.   These lines are filtered and reduced down to the two lines that most likely represent the boundaries of the player's lane.  Then an overlay is applied to the lane between these two lines.   The player is separated from the background using the difference matte created in the beginning and drawn on top of the bowling center and overlays.

This is just a proof of concept video.   Something I threw together in a couple of hours to improve a technology I've been working on for another project.  The overlay is very rough and needs to be cleaned up.   In the future, I'd like to implement this in an application that helps bowlers improve their game.  Eventually tracking the ball path, target, breakpoint, player drift (footsteps), pinfall (pin/score stats), and player form (using human pose estimation).   The precise measurements of the bowling center make it an ideal environment for computer vision technologies.   Distance and size can be measured very accurately.  Bowlers interested in improving their game would be able to analyze dozens of aspects of ball motion (speed/rpm), and their form.  For more casual bowlers, the final application could offer automatic scoring and stats tracking. 

The video below was created by hand using After Effects.  While I was doing this work, I thought  that this kind of work could and should be automated: 
https://www.youtube.com/watch?v=L9-ODWSZeKY

More from this channel