Food and Bioprocess Technology vol:7 issue:6 pages:1783-1792
A geometrical model generator for biological products is presented, which uses X-ray computed tomography images of quasi-axisymmetric biological products as input. It was tested with a dataset of 73 scanned Braeburn apples. For each sample, the generator constructed different cross sections. From these sections, contours were extracted and selected. The contours were expressed as a series of shape descriptors. For this purpose, elliptical Fourier descriptors were used. The obtained frequency distributions were transformed to standard normal distributions. On these transformed distributions, the covariance decomposition algorithm was applied. This algorithm generated new sets of descriptors, which opened up a large range of possibilities for generation of representative shape contours. After reverse transformation of the (generated) descriptor distributions, new contours were obtained from the new descriptors. These new contours were converted to 3D geometrical models of biological products by interpolation and revolving. By comparing the volumes of the generated models with those of the scanned fruit, it was shown that the resulting geometrical models have the same variability as the biological variability in the original dataset. This generator is a fast method, which requires minimal user intervention, and creates 3D models including the biological variability as observed in the scanned fruit. Because these 3D geometrical models are directly available as CAD models, they are useful for numerical modelling of transport phenomena in and around biological products.