Exploiting Problem Domain Knowledge for Accurate Building Image Classification

  • Andres Dorado
  • Ebroul Izquierdo
Conference paper

DOI: 10.1007/978-3-540-27814-6_26

Part of the Lecture Notes in Computer Science book series (LNCS, volume 3115)
Cite this paper as:
Dorado A., Izquierdo E. (2004) Exploiting Problem Domain Knowledge for Accurate Building Image Classification. In: Enser P., Kompatsiaris Y., O’Connor N.E., Smeaton A.F., Smeulders A.W.M. (eds) Image and Video Retrieval. CIVR 2004. Lecture Notes in Computer Science, vol 3115. Springer, Berlin, Heidelberg

Abstract

An approach for classification of building images through rule-based fuzzy inference is presented. It exploits rough matching and problem domain knowledge to improve precision results. This approach uses knowledge representation based on a fuzzy reasoning model for establishing a bridge between visual primitives and their interpretations.

Knowledge representation goes from low level to high level features. The knowledge is acquired from both visual content and users. These users provide the interpretations of low level features as well as their knowledge and experience to improve the rule base.

Experiments are tailored to building image classification. This approach can be extended to other semantic categories, i.e. skyline, vegetation, landscapes. Results show that proposed method is promising support for semantic annotation of image/video content.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer-Verlag Berlin Heidelberg 2004

Authors and Affiliations

  • Andres Dorado
    • 1
    • 2
  • Ebroul Izquierdo
    • 2
  1. 1.School of EngineeringPontificia Universidad JaverianaColombia
  2. 2.Electronic Engineering DepartmentQueen Mary,University of LondonLondonUK

Personalised recommendations