Modelling of Cutting of Pig Carcasses

This thesis describes the methods used to analyse and model the cutting of pig carcasses into the products: back, belly, spine and ribs. Automatic pre-processing methods to segment out the individual products from raw DICOM images are developed. A bit resolution reduction method is presented for producing images that are readable for other software. Image registration method are studied and used for finding the cuttings, namely Thirion’s demons, nonrigid B-spline based free-form deformations and thin plate splines. Image registration gave no success. Insted, a set of ad hoc method were delveloped for finding the cuttings, which turned out to give reasonable results. Statistical shape models are presented and used on the found cuttings. The statistical shape model is based on a 13 eigenvector description spanning 95% of the data space. The outline for a virtual cutting is given. No ground truth is available for the data. Instead, the results are evaluated by visual inspection, since no other possibility exist.