Computer vision is successfully used in sport for purposes of player and object tracking, referee assistance and improved sports broadcasting. But it is not used yet in the steel version of sport darts. Detection of dart locations could be used for automatic point counting and statistic making.
In this thesis we developed two prototypes for detecting and calculating darts locations. Both prototypes use a system of three cameras, which are pointed to the dartboard. In both prototypes we are detecting the colored dart shafts, instead of whole darts. Orientation line is then calculated for each detected shaft. The first prototype uses background subtraction and homography. On the foreground, which represents darts, we search along the orientation line for the location where the dart meets the dartboard. Then, we map the location to the target plane with homography, which gives us the location on the dartboard. The second prototype uses 3D reconstruction. The dart location is found by intersecting the dart orientation line with the plane which represents the dartboard.
Both prototypes are tested with real dart throws. We test the situations where the darts are close and apart, to one another.
|
