Evaluación de técnicas avanzadas de regresión y de características de textura en imágenes de resonancia magnética para determinar parámetros de calidad en productos cárnicos

  1. Ávila Vegas, Mar
unter der Leitung von:
  1. María Luisa Durán Martín-Meras Doktorvater/Doktormutter
  2. María Trinidad Pérez Palacios Doktorvater/Doktormutter
  3. María Teresa Antequera Rojas Doktorvater/Doktormutter

Universität der Verteidigung: Universidad de Extremadura

Fecha de defensa: 23 von März von 2018

Gericht:
  1. Eva Cernadas García Präsidentin
  2. Raúl Grau Meló Sekretär/in
  3. Pablo García Rodríguez Vocal

Art: Dissertation

Zusammenfassung

This Doctoral Thesis focuses on evaluating the prediction of quality parameters of meat products through Magnetic Resonance Imaging (MRI), texture analysis algorithms and advanced regression methods. The use of MRI to determine the quality of meat products is proposed as an alternative and / or complementary technique to the current physico-chemical and sensory analysis. These usual techniques are laborious, costly in time and money, and involve the destruction of meat pieces. MRI is not only a nondestructive technique, but also non-invasive and innocuous. The proposed methodology is based on the acquisition of MRI of loin and Iberian ham by means of low and high field MRI devices, respectively. These MRI are analyzed by using a wide variety of techniques for extracting texture characteristics (gray level coocurrence, fractal analysis, local binary patterns, Wavelet transforms and Gabor filters). Therefore, feature vectors are obtained and then processed by different regression methods (28 regressors) to predict quality parameters by using a realistic validation technique. As well, the study of volumetric structures is carried out through 3D reconstruction of products and the analysis of 3D textures with the aim of extracting additional information which is not obtained by means of the usual techniques. The obtained results confirm the viability of the application of computer vision techniques and MRI to predict quality characteristics of meat products in a non-destructive way.