Kernel based eigenvalue-decomposition methods for analysing ham

AbstractEvery consumer wants fresh ham and the way we decide whether the meat is fresh or not is by looking at the color. The producers of ham wants a long shelf life, meaning they want the ham to look fresh for a long time. The Danish company Danisco is therefore trying to develop optimal storing conditions and finding useful additives to hinder the color to change rapidly. To be able to prove which methods of storing and additives work, Danisco wants to monitor the development of the color of meat in a slice of ham as a function of time, environment and ingredients. We have chosen to use multi spectral images to monitor the change in color. We therefore have to be able to segment the ham into the dierent categories of which the ham consists. These categories include fat, gristle and two dierent types of meat. This segmentation is difficult when using the traditional orthogonal transformation methods, such as PCA, MAF or MNF. We therefore investigated the applicability of kernel based versions of these transformation. This meant implementing the kernel based methods and developing new theory, since kernel based MAF andMNF is not described in the literature yet.

The traditional methods only have two factors that are useful for segmentation and none of them can be used to segment the two types of meat. The kernel based methods have a lot of useful factors and they are able to capture the subtle differences in the images. This is illustrated in Figure 1. You can see a comparison of the most useful factor of PCA and kernel based PCA respectively in Figure 2. The factor of the kernel based PCA turned out to be able to segment the two types of meat and in general that factor is much more distinct, compared to the traditional factor. After the orthogonal transformation a simple thresholding is enough to segment the ham and to detect the color of a type of meat can be done by averaging the pixels that is categorised as that type of meat. Graphs of the change of color in ham as well as more images of the segmentation is included in the article found on
TypeConference paper [Abstract]
Year2010    Month November    pp. 2-3
PublisherThe International Association for Spectral Imaging, IASIM
AddressDublin, Ireland
BibTeX data [bibtex]
IMM Group(s)Image Analysis & Computer Graphics, Geoinformatics