A mobile phone digital image method designed for efficient durum wheat flour characterization

The characterization and counting of foreign bodies and impurities in flour and semolina from wheat are decisive analytical parameters of safety and quality which should be evaluated in the context of efficiency and low cost. Moreover, their recognition and counting are required by the regulatory and quality norms of the sector. International standards such as UNI 10941:2001 could be applied but this has subjective interpretation. In this paper, a new analytical semi-automatic method based on digital images is assessed and proposed. The main features of the method are based on the use of representative sample images that could be taken under controlled illumination conditions with a smartphone. The images were then analyzed using a macro-script adapted to the free software Fiji-ImageJ for image processing. By changing the image color format to grey and its contrast, a threshold intensity, or cutoff, was selected to distinguish foreign bodies from the rest of the product. And the number of particles was counted. Finally, four different fractions of components in the product were recognized which characterized the type of product. The estimated processing time was less than 60 s. The method has been validated against 14 reference samples that were previously studied using the standard UNI 10941:2001. These samples presented low, medium and high particle content, as well as different background colors of the matrix product, from white to yellow. The results were obtained with an average of 6 ROIs taken in different locations of the same digital image. Figures of merit of the procedure such as the biases presented relative differences of less than ±20%, against reference values in the worst case. The reproducibility of the measurements, examining different locations, is better than 30-40% RSD. Likewise, the reproducibility in the same ROI but with short scans in the threshold of selection produced response ranges of less than 30% RSD. These parameters are consistent with those prevalent in the sector and this type of product.