Structured light 3D free form recovering with sub-pixel precision

Fakhr-Eddine Ababsa (1) , David Roussel (1) , Malik Mallem (1)

1 LSC - Laboratoire Systèmes Complexes

This paper deals with the 3D Free form object pose recovering problem that is present in several industrial applications, such as online quality control in production as well as image systems for assembly/welding, augmented reality and robotics. The method presented here uses a structured light based vision system to reconstruct several accurate local 3D patches of the objects. A robust subpixel method for image features detection has been developed in order to increase 3D reconstruction accuracy. The Curvature method is then used to compute a geometric invariant "footprint" to discriminate reconstructed patches, which allows to match them with the object's model. Pose recovering is performed by using the prediction - verification hypotheses paradigm. Some experimental results are given to show the efficiency of the proposed solution when applied to a complex free form object.

Mots clés: Free form object; structured light; sub-pixel detection; 3D recovery; Matching; Geometric Invariant; recognition by indexation

Domaines: Traitement du signal et de l'image [eess.SP]; Traitement du signal et de l'image [eess.SP]

Fakhr-Eddine Ababsa, David Roussel, Malik Mallem. Structured light 3D free form recovering with sub-pixel precision. Machine Graphics & Vision, 2003, 12 (4), pp.453 --476. ⟨hal-00376699⟩