Recognition and reconstruction of partially occluded objects

Abstract - A new automatic system for the recognition and reconstruction of rescaled and/or rotated partially occluded objects is presented. The objects to recognize are described by many 2D views and each view is occluded by half-planes with different slopes. The remaining parts (linear cuts) and the whole object views are then stored in a database. To establish if a region R of an input image represents an object possibly occluded, the system generates a set of linear cuts of R and compare them with the elements in the database. Each linear cut of R is associated to the most similar database linear cut. R is recognized as an instance of the object O if the most of the linear cuts of R are associated to a linear cut of views of O. In the case of recognition, the system selects the region cut and the correspondent view cut C (O) whose log-polar transforms match as best as possible and uses them to reconstruct the whole shape of R. The scale factor and orientation in image plane of R with respect C (O) are determined.