Necessary views for a coarse representation of a scene

Presents an approach to a typical task of active sensing, namely how to position a camera in order to acquire the necessary geometric information to determine the maximal height of each object in the scene. The possible camera directions are limited to assure the precise data acquisition. First, the top view of the scene is obtained, which gives the arrangement of the objects and provides the geometric information necessary to compute the azimuth values for the side views. We have designed a method which determines the minimal number of side views that are sufficient to accomplish the task. Experimental results demonstrate the utility of the proposed method, which can be used in industrial environments, in particular for tasks related to robot grasping