Back to search

National Technical University of Athens (NTUA)

Διαφορική γεωμετρία και θεωρία μεταβολών στην αναγνώριση προτύπων και την ψηφιακή ανάλυση εικόνας, με πρωτότυπες εφαρμογές στη βιοϊατρική και την αρχαιομετρία

Abstract

dc:description

In the present thesis, new data similarity quantities are defined and associated data grouping/clustering algorithms are developed, exploiting a) notions and new approaches from the field of Calculus of Variations and b) representations of the data and of their congruence as differentiable manifolds in two, three and four dimensions.The three dimensional case concerns the mathematical formulation and resolution of the automatic fragmented objects reconstruction problem. The formulation of the problem is based on the quantification of the diversification of the adjacent fracture faces by means of error functionals defined over three, two and one dimension. The resolution of the reconstruction problem assembles the variational solutions of the constrained maximization of the geometric diversification quantities in a hierarchy consistent with the joint validity of these diversification extremes. This hierarchy is proved to be unique up to isomorphisms and, therefore, it governs the actual automatic fragmented objects reconstruction algorithm. As a part of an integrated automatic wall-paintings reassembly system, this algorithm has been applied for the determination of reconstructed islands among the fragments of prehistoric frescos excavated at Tyrins and held in the Prehistoric Collection of the National Archaeological Museum in Athens.The two and four dimensional manifolds cases correspond to the determination of congruence relations between curves up to elastic deformations and deformations of curves' implicit functional representations, respectively.Specifically, in the first case, the differentiable deformations of 2D bodies are expressed in terms of curvilinear coordinates and the associated system of PDEs that govern the deformations is determined. Moreover, the solution of the deformation PDEs is associated with an equivalent succession of morphological operators. Due to the form of the selected curvilinear coordinates system, the 2D deformations are decoupled as a combination of elastic deformation with local dilations/erosion. By means of the morphological operator that corresponds to the elastic deformations, two procedures are developed, which determine the neutral line of each such deformation and reconstruct undeformed versions of 2D bodies, straightening the spotted neutral line. On the basis of these procedures, a computer system has been developed that reconstructs undeformed versions of bodies from images of them in arbitrary deformation instances. This system, integrated with a properly formulated curves classification method has been used for the automatic classification of parasites from their microscopic images.Finally, in the 3rd subject of the thesis, congruence relations of curves are formulated so as to correspond to embeddings in level-sets of the same functions R4 → R. It is shown that such embeddings include every non-Euclidean, differentiable congruence of curves in R2, also being minimal in terms of the dimensions of the representation of this congruence. Then, projecting the generalized shape operator in the tangent space of the corresponding level-sets and using the generalized Stokes theorem, we have formulated the optimal embedding of a given curve segment in the level-sets of an implicit family of prototypes, in terms of the deviation of the level-sets curvature. It is shown that the minimization of this deviation implies the minimization of the geodesic distance in the 4-manifold formed by the level-sets of a certain implicit family of prototypes, at the same time keeping the fitting error invariant under Euclidean transformations. By demanding that the fitting error remains stationary on a submanifold of the level-set that includes the best fitting prototype, two complementary fitting procedures have been developed: a) the first one determines the best fitting congruence between an implicit prototype and a given "data" curve and b) the second one determines the prototype that optimally aligns more than one "data" curves along it. These procedures are the basis of an integrated computer system that determines the minimal number of prototype curves, which optimally match the maximum possible portion of a given set of curve segments. Application of this system to the analysis of the contours of the drawn figures of Mycenaic frescos, kept in the National Archaeological Museum in Athens, indicated that the wall-painting "Lady of Mycenae" have been constructed using advanced geometric guides that correspond to 2 hyperbolae and 2 linear spirals. Moreover, it has been rejected the hypothesis that a second wall-painting with stable contours could have been drawn using any guide in the form of conics or linear (Archimedes), exponential or spiral of revolution.

Degree

thesis:*
Grantor dc:publisher
National Technical University of Athens (NTUA)
Year dc:date
2015

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Arabadjis, Dimitrios
  • Αραμπατζής, Δημήτριος

Subjects

dc:subject × 22

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/38115

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Arabadjis, Dimitrios; Αραμπατζής, Δημήτριος. Διαφορική γεωμετρία και θεωρία μεταβολών στην αναγνώριση προτύπων και την ψηφιακή ανάλυση εικόνας, με πρωτότυπες εφαρμογές στη βιοϊατρική και την αρχαιομετρία. National Technical University of Athens (NTUA), 2015. http://hdl.handle.net/10442/hedi/38115