Kovalenko M. Ontological model based automatic video annotation information technology

Українська версія

Thesis for the degree of Candidate of Sciences (CSc)

State registration number

0415U004454

Applicant for

Specialization

  • 05.13.06 - Інформаційні технології

28-08-2015

Specialized Academic Board

Д 41.052.01

Odessa National Polytechnic University

Essay

The Mixture of Gaussians based moving object detection method was improved by using a texture similarity hypothesis and the normalized pixel intensity cross-correlation. An ontology based video content representation model is proposed, which uses ontological rules. An improvement on the ontological inference method is proposed by introducing the temporal logic. A Bayesian network based method to close the semantic gap is proposed. A branch-and-bound based method to learn the optimal Bayesian network structure was improved by adding the structural constraints. Based on the proposed model and methods the automatic video annotation information technology's software was developed.

Files

Similar theses