Innovative automated landmark detection for food processing: The Backwarping approach

8 pages, 4 figures, 2 tables
The shape of an object can be described by a finite number of points called landmarks. Nowadays, there are no systems available for the automated landmarks detection. However, the implementation of such method would be of elevated interest in the food industrial processing. The evaluation of cattle carcass and fish quality requires the time-consuming and manual positioning of landmarks, with still too subjective results. The aim of this work is the application of an innovative algorithm, called backwarping, for the automated positioning of landmarks onto the beef carcass and sea bass profiles. This algorithm combines the automated extraction of the outlines with the thin-plate spline interpolation algorithm. The proposed approach is applied to two very different cases in order to stress the high performing versatility of the procedure. We firstly carried out a calibration phase where the landmarks were manually placed. Then we applied the automated procedure for comparison. The errors in the automated landmarks positioning has been always lower than 3 % and the percentage standard error of prediction was always lower than 1.5 %. The approach for both study cases showed its feasibility in being easily extended to the processing of a diversified variety of food products, such as poultry, bakery or >pasta.> Moreover, due to its versatility, the technique could be also applied within the zoological and biomedical fields, given the obtained high levels of accuracy in the automated landmark positioning. © 2013 Springer Science+Business Media New York
This study was supported by the project “Consorzio produzione carne bovina” (ANABIC) funded by ItalianMinistry of Agricultural, Alimentary and Forestry Politics. Jacopo Aguzzi is a Research Fellow within the Ramon y Cajal Program (MICINN)