Pointing tasks frequently require highly accurate tracking of the user's point of aim. An example is tracking the aim of a user within a video game environment, or calculating the orientation of a mobile robot using data from an onboard camera. Many current methods for tracking user aim are not very accurate, and only capable of tracking relative motions of the pointer. Recognizing this, a system for interacting with a visually projected firearm target practice simulation is presented. An invisible laser is mounted on a training firearm and pointed at a video projection screen. The location of the laser dot on the projector screen is tracked by a camera. Several non-parametric error correction techniques are used to compensate for the distortion of the resulting image in real time and allow for accurate tracking of the dot location. Several targets are generated, and the user engages them by aiming along the standard weapon sights and pulling the trigger. The accuracy of system is evaluated by aiming the weapon at several points not used in the error correction process, and averaging the error between the calculated and actual aimpoints at these locations. This metric shows that the system produces an average normalized error of 0.0013 across the projected field, which corresponds to an aiming error of 3.4 inches at 100 yards.