Automatic procedure to contactless and non-destructive quality evaluation of fruits and vegetables through a computer vision system

Quality rating of fruit and vegetable is currently accomplished by non-destructive and subjective sensory evaluation or by objective and destructive analytical techniques. There is a strong need of objective non-destructive contactless quality evaluation systems to monitor fruit and vegetable along the whole supply chain even after the packaging phase. Computer vision systems (CVS) can satisfy this request through a combination of image processing and machine learning techniques. Color correction, foreground extraction, recognition of relevant colors, features extraction and selection, classification and regression are components that must be developed to achieve an efficient and robust quality evaluation. Machine learning can automatically configure the parameters normally set by operators in ordinary CVS and can automate the adaptation of the system to the specific product at hand making easier its extension to new applications. The paper describes the results obtained by an innovative flexible and self-configuring CVS: the proper integration of machine learning methods in the CVS proved successful in non-destructive contactless global quality evaluation and internal parameters estimation on different products with minimal efforts during the design and configuration phase. In detail, models previously published, were applied to evaluate the quality level of fresh-cut radicchio, Iceberg lettuce head and fresh-cut, and table grape bunches commercially available. The critical tasks involved in each component of the CVS and their relevance for the final effectiveness of the system will be described. The solutions provided by the machine learning techniques pointed out to the efficacy and flexibility of the resulting system.