This research deals with the design of a vision system for a specified task facilitated by sensor-placement techniques that efficiently compute the set of viewpoints from which a specified surface target is not occluded by any object in the environment and from which the entire target lies within the pyramidal field view of the sensor. The target, which cannot be properly viewed by one sensor, is partitioned into a minimal number of components each of which is covered by one sensor. The following issues are addressed: acceptable locations for triangulation range sensors; appropriate camera parameters (image plane, lens, and aperture diameter), given an approximate sensor location; sensor region for targets composed of curved surfaces; and the best viewpoint for a given task.
