Robust Extraction of 3D Structures by Fusion of Intensity-Based and Contour-Based Junction Features

This paper describes a new method for validating and classifying 3D junctions by combining detected intensity-based junctions and contour-based junctions. The resulting algorithm is divided in four steps: (i) the pairing of junctions according to proximity criteria, (ii) the matching of brancheslsegments of paired junctions, (iii) the validation of isolated junctions according to paired junctions and finally, ( iv) the validation of b r a n c h e s l s e g m e n t s t o c o m p l e t e t h e j u n c t i o n characterization. Preliminary experimental results are presented which show the effectiveness of the method to infer in a robust manner the description of 3D objects.