next up previous
Next: Tracking Objects Up: Initial Search Previous: Localization using Evidence Aggregation


Initial Localization of Ball and Obstacles

Finding the ball is of essential importance for successful play. Although it seems to be an easy task, detecting the ball clearly demonstrates the difficulties of computer vision. Figure 3 shows some images of the ball. Its appearance varies greatly in size, shape, and color. Furthermore, it can be easily confused with robot markers. The detection of the ball can be summarized in three steps: (a) determine all clusters of the ball's color class and their sizes, (b) discard all clusters for which no transition to the floor can be found, and (c) choose the biggest cluster. Obstacles are detected as large clusters of the dark obstacle color. If clusters are close to each other, they are merged.

Figure: The ball's appearance varies in color and shape and is similar to robot markers.
\begin{figure}\begin{center}
\epsfig{file= baelle.eps,height=0.111\hsize}\hspace...
...= colorMarker.eps,height=0.111\hsize}\vspace*{-5ex}
\end{center}\par\end{figure}



Sven Behnke 2001-11-01