3D object recognition through a size function resulting from an invariant topological feature. (December 2021)

• Record Type: Journal Article
• Authors: Alaoui Mhamdi, Mohammed Ayoub; Ziou, Djemel
• Is Part Of: Pattern recognition. Volume 120(2021)
• Publication Date: 2021-12
• Subjects: 3D Object description -- Object recognition -- Object categorization -- Shape classification; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2021.108131
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: The 3 D object is described by a size function. The size function resulted from a measure function that encompasses invariance to affine transformations, articulations, and torsion. The proposed method tackles with the problem of partial matching of 3D objects. Abstract: In this paper, a critical points based descriptor for 3 D objects recognition is presented. It is based on the topological invariant provided by the critical points of the 3 D object. The critical points and the links between them are represented by a size function resulting from a measure function that captures the surface displacement along the 3 D object, and that encompasses invariance to affine transformations, articulations and torsions. In order to tackle the problems of partial matching of the 3 D objects, a well-suited metric learning method is used to weight the matchings according to their relevance. The proposed method's performance was validated by different collections of 3 D objects. The obtained scores are favorably comparable to the related work.