Computer vision system for image analysis of visible fat in dry sausages

The research aimed to investigate the effect of the raw material origin (5 producers used meat bought from the market and 5 used meat from pigs bred on their farms) and of the brand (10 salami factories) on fat/lean ratio in Lucanian dry sausages. The sausage slices were photographed and either used for a consumer test or analysed with the Computer Vision System to evaluate the fat/lean area ratio using image binarization. This analysis converts an image of up to 256 grey levels to a black (lean) and white (fat) image. 349 male and 394 female consumers (n=743) were asked to look carefully at the 6 sausages slices in the photo and to identify the slices they considered unacceptable (each assessor was allowed to choose zero to six slices) and the slice they preferred (each assessor was allowed to choose only one out of six slices). Data gathered from the preference test were analyzed to determine statistical significance based on the frequency of each response (chi-squared one sample test) where the expected frequency was 16.6%. Data concerning acceptability were analyzed using chi-squared for k samples with an expected frequency of 50% (acceptable vs not acceptable). The visible fat/lean ratio of the sausages slices was higher (P<0.001) in the product from salami factories without pig farm. The fat percent- age was highly variable (P<0.001) among the sausages of each salami factory, although this heterogeneity can be considered a typical characteristic of these products. On the whole, the prod- ucts the consumer considered unacceptable had a higher fat percentage (P<0.001). Consumers (about 70%) preferred the two leanest products (P<0.001). Female consumers showed a higher preference for lean products as compared with male (P<0.001).