Computer Vision Technology for Quality Monitoring in Smart Drying System

Abstract

Drying is one of the most viable and effective preservation technologies to improve the shelf-life of foods. Carrots are among the most consumed vegetables, owing to their nutritional profile as well as their wide use in dried foods, ready-to-eat and ready-to-use convenience products like snacks, meals, and soups. As for the dried products, the quality of produce depends on the timely recognition of the dehydration state. Traditional off-line analyses in combination with drying rates to identify the end-time of the process can fail in identifying process discrepancies and avoiding product degradation. The use of computer vision (CV) as a Process Analytical Technology (PAT) tool in the drying system can be of interest to monitor the drying process and product quality. The objective of this study was to study the drying behavior of carrot slices during drying at 35 °C for 36 h using a smart dryer augmented with computer vision system and load cell. The system developed was effective in measuring the weight, size, and color of the untreated (control) and pre-treated (blanched) carrot slices along the drying time. The image analysis and the weight loss of the slices enabled the prediction of relative moisture content (MC) using linear and thin-layer (Newton-Lewis) models in comparison. The applicability of the models was further evaluated by use of different pretreatments (i.e. blanched at 90 °C for 2 min or not blanched). The results showed promising prediction capability for the linear models, which was independent of time with a Root Mean Square Error (RMSE) similar to the thin-layer models, an adj. R2 > 0.99 as well as both Mean BIAS Error (MBE) and reduced Ȥ2 tending towards zero. The blanching treatment affected the model parameters but negligibly affected the model performances.