In the sport of competitive water skiing, the skill of a human boat driver can affect athletic performance. Driver influence is not necessarily inhibitive to skiers, however, it reduces the fairness and credibility of the sport overall. In response to the stated problem, this thesis proposes a vision-based real-time control system designed specifically for tournament waterski boats. The challenges addressed in this thesis include: one, the segmentation of floating objects in frame sequences captured by a moving camera, two, the identification of segmented objects which fit a predefined model, and three, the accurate and fast estimation of camera position and orientation from coplanar point correspondences. This thesis discusses current ideas and proposes new methods for the three challenges mentioned. In the end, a working prototype is produced.