AN OPTIMAL SENSING STRATEGY FOR RECOGNITION AND LOCALIZATION OF 3-D NATURAL QUADRIC OBJECTS

Active sensing aims at achieving a goal-directed collection of data critical for sensing goals by controlling sensor configurations and poses. Active sensing thus requires an optimal sensing strategy for controlling sensor configurations and poses in such a way as to minimize data collection operations necessary for achieving sensing goals. This paper presents an optimal sensing strategy of an optical proximity sensor system engaged in the recognition and localization of 3-D natural quadric objects. The optimal sensing strategy consists of the selection of an optimal beam orientation and the determination of an optimal probing plane that compose an optimal data collection operation known as an optimal probing. The decision of an optimal probing is based on the measure of discrimination power for a cluster of surfaces on a multiple interpretation image (MII), where the measure of discrimination power is defined in terms of a utility function computing the expected number of interpretations that can be pruned out by a probing. This paper also presents an object representation suitable for active sensing based on a surface description vector (SDV) distribution graph and hierarchical tables. Experimental results are shown.