Spatially Resolved Spectroscopy for Non-Destructive Quality Inspection of Foods

Consumer requirements for food quality in Europe have been changing in the past decades. Qualified food products are required to be not only safe, nutritious and healthy, but also tasty and produced in a sustainable way. The fact that the food taste is actually determined by both chemical composition and texture during mastication makes that controlling and evaluating the microstructure which relates to texture along the production process and in the final food product have become important. This implies that the microstructural quality of each individual final food product and even by-products has to be tested to guarantee maximal customer satisfaction. However, current methods for food microstructure assessment are mostly based on microscopy techniques which are time consuming, destructive and require complex and expensive equipment. These microscopy techniques, therefore, are not suitable for use in automatic process operation in the food industry and for testing the microstructure quality of individual products.Vis/NIR spectroscopy has been extensively utilized in research in combination with multivariate statistical methods for non-destructive quality inspection of numerous agro-foods. The basic principle of this spectroscopic measuring technique is making use of the interaction of Vis/NIR light with the sample medium which involves both light scattering, caused by inhomogeneities in refractive index inside the sample, and light absorption by chemical components. So far, Vis/NIR spectroscopy has mostly been used for prediction of the chemical composition by means of multivariate calibration models based on Beer s law. As the scattering phenomena related to microstructure of the measured samples increase, the path lengths travelled by the propagating light cause deviations from Beer s law and thus are typically filtered out during construction of these multivariate models. The inability to separate variation in the Vis/NIR spectra caused by chemical composition and microstructure limits the use of conventional Vis/NIR spectroscopic measurement techniques for quantifying the microstructure of agro-foods and hence their textural quality attributes.To promote fast and non-destructive microstructure assessment in the agro-food industry, this PhD research aimed at acquiring information on the scattering and absorption properties of agro-food samples. For this purpose multiple conventional Vis/NIR spectroscopic measurements at different spatial locations from the illumination point, spatially resolved spectroscopy (SRS), were combined. The absorption and scattering information represented by the absorption coefficient µa and the reduced scattering coefficient µs spectra were then used to evaluate the quality attributes of the studied agro-foods related to their microstructure and chemical composition. First, a setup for SRS measurements in reflectance mode in the 500 1000 nm range was elaborated and a measurement procedure was established and calibrated. The accuracy of the SRS setup in acquiring the spatially resolveddiffuse reflectance and the inverse procedure for optical properties estimation involving a metamodel constructed with Stochastic Kriging were evaluated for liquid optical phantoms with known optical properties. This optical properties estimation procedure was then applied for estimating the optical properties of three food samples with increasing microstructure complexity: sugar foams as model systems for aerated sugar containing foods, osmo-air-dried apple slices and fresh Braeburn apples subjected to long-term storage and shelf-life. The estimated scattering properties were then used for non-destructive estimation of the microstructure parameters of these measured samples and compared with those measured by X-ray micro-CT technique. Finally, these were also related to the quality attributes as measured by the reference destructive methods.The obtained results indicated that the metamodeling method was well capable of both predicting the spatially resolved diffuse reflectance of samples with known optical properties measured by the SRS setup and acquiring the estimated optical properties of the liquid optical phantoms with relative prediction errors of 10% for µa and µs . Clear optical properties-microstructure-quality relations were observed for the sugar foams and osmo-air-dried apple slices. This confirms the potential of using scattering information for assessing the microstructure and textural quality attributes of these food samples and for monitoring and optimization of food processes. For fresh Braeburn apples, the absorption information clearly indicated the changes in chemical composition of the apples (e.g. chlorophyll content) during long-term storage and shelf-life. These outcomes could be used for monitoring the changes in quality of the apples during long-term storage and shelf-life. However, the prediction performance for quality attributes related to texture and chemical components (firmness and soluble solids content) obtained by SRS did not outperform classical single-point Vis/NIR spectroscopy. Thanks to its fast estimation of the optical properties with acceptable accuracies, the metamodeling approach employed in this research for estimating optical properties of agro-food samples holds potential for on-line applications for non-destructive food microstructure, and hence texture, control and assessment or food process optimization and monitoring.status: publishe