Procesado de imágenes por técnicas de multiplexado. Aplicación al reconocimiento de objetos tridimensionales

  1. Esteve Taboada, José J.
Dirigida por:
  1. Javier García Monreal Director
  2. Carlos Ferreira García Director

Universidad de defensa: Universitat de València

Fecha de defensa: 09 de julio de 2003

Tribunal:
  1. María Josefa Yzuel Giménez Presidente/a
  2. Pascuala García-Martínez Secretaria
  3. María Sagrario Millán García-Varela Vocal
  4. Gabriel Cristobal Pérez Vocal
  5. Santiago Vallmitjana Rico Vocal
Departamento:
  1. Òptica i Optometria i Ciències de la Visió

Tipo: Tesis

Teseo: 89639 DIALNET lock_openTDX editor

Resumen

Several applications in the fields of optical pattern recognition and image processing using multiplexing techniques have been presented. Using wavelength multiplexing, we have improved the detection capabilities of a white-light optical correlator that now detects simultaneously different scaled reference objects up to a magnification factor equal to 2. The color of the correlation peak indicates us directly the scale factor of the detected reference object. As an additional application, the same experimental setup permits to obtain, in real time, the different wavelet components of an input function, in the same spatial region, and each one codified in a different interval of wavelengths. The interval of wavelengths in which each wavelet component appears can be selected by changing some characteristics of the system. We have also presented a new full-optical system to localize a reference object in the three-dimensional space. The system multiplexes each different depth position in a different wavelength, and so, the color of the obtained correlation peak permits to know without confusedness the exact localization of the object in the three-dimensional scene. Additionally, using the spatial multiplexing technique, we have presented a new full-optical system to recognize three-dimensional objects. The detection process can be performed in real time, because no digital processing of the input image is needed. The optical system, that is based on the Fourier transform profilometry technique, can be implemented using a joint-transform correlator or a classical convergent correlator. We have also extended the recognition capabilities of the new introduced technique to obtain rotation-invariant three-dimensional object recognition, and scale-invariant three-dimensional object recognition. All the results we present have been obtained experimentally, and demonstrate the good performance of all the applications and proposed systems.