TEXTURISE: Development of applications to predict food texture using non-destructive imaging technologies and artificial intelligence.
Food texture is an important quality attribute and, therefore, ensuring optimal texture is essential for the food industry. This can be achieved through the development of control tools capable of determining food texture at the end of the process and identifying products with defective textures. In this regard, computer vision and artificial intelligence have been used in recent years to automate many industrial processes, and the food industry is no exception. However, their use for predicting the sensory quality of food is still emerging.
Objectives
The objective of TextuRise-2 is to investigate the use of computer vision and artificial intelligence to determine sensory texture in foods affected by texture-related problems, specifically to assess texture in high-moisture extrudates and detect pastiness defects in dry-cured ham, as well as to develop software to facilitate their use or industrial implementation.To this end, two non-destructive imaging technologies will be used: 1) colour cameras operating in the visible range, which provide information on the visual appearance of the product; and 2) hyperspectral cameras, which can provide information on the chemical composition of the product. These images will be used to design new algorithms to predict the anisotropy index and sensory fibrousness level of alternative-protein extrudates, as well as the proteolysis index and sensory pastiness level of dry-cured ham slices.Prediction performance using classical approaches, based on a sequence of preprocessing, segmentation, feature extraction and machine-learning modelling stages, will be compared with approaches based on deep learning.The TextuRise-2 project will investigate which acquisition technology (RGB images or hyperspectral data) provides more information for predicting food quality attributes. This will generate knowledge about how an inherent food property such as texture can be perceived through imaging. The integration of different sources of information for predicting sensory quality attributes in the food products under study will also be investigated.Finally, two software applications will be developed, one for each product studied, to facilitate their use in research laboratories and/or their implementation at an industrial level.
Project
/research/projects/desenvolvemento-de-aplicacions-para-predicir-a-textura-dos-alimentos-mediante-tecnoloxias-de-imaxe-non-destrutiva-e-intelixencia-artificial
<p>Food texture is an important quality attribute and, therefore, ensuring optimal texture is essential for the food industry. This can be achieved through the development of control tools capable of determining food texture at the end of the process and identifying products with defective textures. In this regard, computer vision and artificial intelligence have been used in recent years to automate many industrial processes, and the food industry is no exception. However, their use for predicting the sensory quality of food is still emerging.</p><p>The objective of TextuRise-2 is to investigate the use of computer vision and artificial intelligence to determine sensory texture in foods affected by texture-related problems, specifically to assess texture in high-moisture extrudates and detect pastiness defects in dry-cured ham, as well as to develop software to facilitate their use or industrial implementation.To this end, two non-destructive imaging technologies will be used: 1) colour cameras operating in the visible range, which provide information on the visual appearance of the product; and 2) hyperspectral cameras, which can provide information on the chemical composition of the product. These images will be used to design new algorithms to predict the anisotropy index and sensory fibrousness level of alternative-protein extrudates, as well as the proteolysis index and sensory pastiness level of dry-cured ham slices.Prediction performance using classical approaches, based on a sequence of preprocessing, segmentation, feature extraction and machine-learning modelling stages, will be compared with approaches based on deep learning.The TextuRise-2 project will investigate which acquisition technology (RGB images or hyperspectral data) provides more information for predicting food quality attributes. This will generate knowledge about how an inherent food property such as texture can be perceived through imaging. The integration of different sources of information for predicting sensory quality attributes in the food products under study will also be investigated.Finally, two software applications will be developed, one for each product studied, to facilitate their use in research laboratories and/or their implementation at an industrial level.</p> - PID2024-156581OR-C22 - Eva Cernadas García
projects_en