Quality assessment of commercial bread samples based on breadcrumb features and freshness analysis using an ultrasonic machine vision (UVS) system

This paper presents the use of a in situ developed ultrasonic machine vision system for quality parameter extraction of breadcrumb features and freshness. An image processing technique has been used for breadcrumb analysis on collected digital images of various bread samples while an ultrasonic assessment technique has been used for quantification of the freshness of various bread samples. Various threshold methods (isodata, Otsu, minimum error, moment preserving and fuzzy method) have been implemented and compared with the proposed method to segment breadcrumbs from collected digital bread images. Threshold performance was assessed by two important criteria such as uniformity and busyness (arrangement of a pixel to its neighborhood pixels) of the binary versions of input breadcrumb sample images. Quality parameters were computed for each optimal threshold on 500 digital images of bread slices. Other important quality parameter of bread is the outline of its brown color section, which corresponds to the appropriate baking stage. Slight variations in threshold lead to substantial variations in crumb feature values, with cell uniformity, void fraction, intensity and entropy calculation showing more sensitivity than others. Propagation delay and attenuation in the received acoustic signal have been calculated for stiffness and firmness evaluation. A second order relationship has been observed between the storage time and stiffness of the various bread samples. The proposed method is very efficient in the sense of quality parameter calculations. Although some of the previously reported methods showed a relatively higher amount of busyness than other methods, the reported method performs well on images with large void areas.