Full text
| |
| Author(s): |
Juliana Monteiro Balage
Total Authors: 1
|
| Document type: | Doctoral Thesis |
| Press: | Pirassununga. |
| Institution: | Universidade de São Paulo (USP). Faculdade de Zootecnica e Engenharia de Alimentos (FZE/BT) |
| Defense date: | 2017-11-14 |
| Examining board members: |
Saulo da Luz e Silva;
Douglas Fernandes Barbin;
Marina de Nadai Bonin;
Mario Luiz Chizzotti;
Marco Antonio Trindade
|
| Advisor: | Saulo da Luz e Silva |
| Abstract | |
Increasingly, industry requires real-time methods for quality control of fresh meat in order to improve production efficiency, ensure product homogeneity and meet consumer expectations. In the present work, the hyperspectral image was used to evaluate the quality of Nellore beef with emphasis on tenderness and characteristics related to it, and also the construction of distribution maps to observe the variability of these characteristics between and within samples. To investigate whether the use of different muscle groups increases the variability of the reference values, improving tenderness prediction and classification models, samples from Longissimus (94) and B. femoris (94) of Nellore cattle were used. To investigate whether the selection of the region of interest (ROI) in the image at the exact location where the shear force cores were collected improves tenderness prediction and classification models, samples from Longissimus muscle were used (50). After image acquisition (1,000 - 2,500 nm), each sample was evaluated following traditional methodology for shear force, dry matter, crude protein, lipids and sarcomere length. The spectral and spatial data were analyzed by chemometric techniques and PLSR and PLS-DA models were constructed. Regarding the approach with different muscles, the data were modeled separately to avoid that phenomena due to muscle differences were mistakenly attributed to the characteristics investigated. Nevertheless, samples from Longissimus with unacceptable tenderness were classified with sensitivity = 87% and tender samples from B. femoris with sensitivity = 90%, both in the external validation. Regarding the ROI selection, the classification models using local ROI presented better performance than the ROI models of the whole sample (external validation sensitivity for the tough class = 33% and 70%, respectively). However, the more general model had better performance in the tenderness distribution maps, with 72% of predicted images correctly classified. (AU) | |
| FAPESP's process: | 15/00293-3 - Beef quality evaluation using near infrared hyperspectral imaging |
| Grantee: | Juliana Monteiro Balage |
| Support Opportunities: | Scholarships in Brazil - Doctorate |