Artificial neural networks and electron microscopy to evaluate the quality of fruit and vegetable spray-dried powders. Case study: Strawberry powder

Abstract

The increasing trend of consumer awareness regarding balanced nutrition, methods of food preparation and processing as well as an increased attention to quality of products by customers have resulted in the phenomena of searching for new methods of improving the quality of product and automatization of technological processes.The research was based on image recognition of qualitative classes of powders obtained by the use of the process called industrial spray-drying. Structures of spray dried strawberry juice were analyzed using scanning electron microscopy and a digital camera. To this end, Gray-Level Co-occurrence (GLCM) and Laws texture analysis were used. The final outcome obtained as a result of the above tests demonstrated that the analysis of quality properties of texture made it possible to effectively distinguish research material consisting of various spray-dried powders of strawberry juice. The analysis of Artificial Neural Networks showed that there are adequate models in terms of the lowest error value RMS (Root Mean Square) and effectiveness classifying wrong structure of particles in assessment of the qualitative strawberry powders.The analysis of key components demonstrated that texture parameters correlate well with color feature in the technique called computer image analysis. More strongly correlated variables were presented in GLCM texture parameters in scanning electron microscopy. A vital correlation depended on recognition of microstructure of particles with regard to shape such as Ferret factor, roundness factor or circularity factor in electron microscopy.These results confirm that recognition of qualitative classes of powders can constitute the basis for quick evaluation of the quality of fruit powders in industrial spray drying.