Feasibility of computer vision as Process Analytical Technology tool for the drying of organic apple slices

Abstract

Quality and sustainability of product depend on the cumulative impacts of each processing step in the food chain and their interplay. Various research studies evidenced that many drying systems operate inefficiently in terms of drying time, energy demand (e.g. fossil fuels), raw material utilization and resulting product quality. Moreover, not all conventional drying processes are allowed in the organic sector. In recent years, non-invasive monitoring and control systems have shown a great potential for improvement of the quality of the resulting products. Therefore, intelligent processes enabling simultaneous multifactorial control are needed to ensure high value end products, improve energy and resource efficiency by using microcontrollers, innovative and reliable sensors and embracing various areas of research and development (e.g. computer vision, deep learning, etc.). The objective of this study was to evaluate the feasibility of computer vision (CV) as a tool in development of smart drying technologies to non-destructively forecast changes in moisture content of apple slices during drying. Samples were subjected to various anti-browning treatments at sub- and atmospheric pressures, and dried at 60°C up to a moisture content on dry basis (MCdb) of 0.18 g g-1. CV-based prediction models of changes in moisture content on wet basis (MCwb) were developed and promising results were obtained (R2P>0.99, RMSEP=0.011-0.058 and BIASP<0.06 in absolute value), regardless of the anti-browning treatment. The proposed methodology lays the foundations for a scale-up smart-drying system based on CV and automation.