Spatial Object Perception from an Image

In this paper we address the problem of finding the spatial position and orientation of an object from a single image. It is assumed that the image formation process and an object model are known in advance. Sets of image lines are backprojected and constraints on their spatial interpretations are derived. A search space is then constructed where each node represents a space feature with a model assignment. Next, a hypothesize-and-test recognition strategy is used to select a solution, that is to determine six degrees of freedom of a part from a set of features. Finally we discuss the efficiency and the reliability of the method.