Detection and Recognition of Non-Occluded Objects using Signature Map
Sang‐Bum Park, Youngjoon Han, Hernsoo Hahn · 2007
Abstract:- For constructing a flexible bin picking system where parts can be provided with arbitrarily stacked in a workspace, detection and recognition of non-occluded objects are essential process. For implementing such process, this paper proposes a new algorithm which determines whether an object is occluded or not and at the same time which object it is in the DB of object models. It is based on a signature map which is constructed by detecting the objects in an input image and drawing the signatures of the whole image with reference to individual objects. Thus the number of signature maps is equal to the number of the objects detected in the input image. A signature map shows the outer contour and inside edge features. Occlusion by other objects appears as distortions in the outer contour of the signature map. The inside edge features are used for discerning the objects having the same outer contour by different inside shape. To make the manipulator pick up a selected part, a pose estimation method for elliptical objects is also proposed. The performance of the proposed algorithm has been tested with the task of picking the top or non-overlapped object from a stack of arbitrarily located objects. In the experiment, a recognition rate of 98 % has been achieved.