A set of camera selection templates has been developed and applied to a twelve-camera inward-looking distributed smart camera network mounted on the vertices of an icosahedral frame. These templates, designed using the fundamentals of self-organization, consist of simple rules based on a local (camera) level metric - a measure of the quality of detection of the target-of-interest for a given camera node based on a measurable target parameter. The effectiveness of the camera selections, based on the local metric, is analyzed through the performance of a global (system) level metric. The camera selection templates are shown to maintain a desirable global metric performance, while using a subset of the total available cameras. This holds true even when a single occluding target is introduced into the system.