Near infrared hyperspectral imaging for the evaluation of endosperm texture in whole yellow maize (Zea maize L.) Kernels

Simple item page
dc.contributor.authorManley M.
dc.contributor.authorWilliams P.
dc.contributor.authorNilsson D.
dc.contributor.authorGeladi P.
dc.date.accessioned2011-05-15T15:56:50Z
dc.date.available2011-05-15T15:56:50Z
dc.date.issued2009
dc.description.abstractNear infrared hyperspectral images (HSI) were recorded for whole yellow maize kernels (commercial hybrids) defined as either hard, intermediate, or soft by experienced maize breeders. The images were acquired with a linescan (pushbroom) instrument using a HgCdTe detector. The final image size was 570 × 219 pixels in 239 wavelength bands from 1000 to 2498 nm in steps of approximately 6.5 nm. Multivariate image cleaning was used to remove background and optical errors, in which about two-thirds of all pixels were removed. The cleaned image was used to calculate a principal component analysis (PCA) model after multiplicative scatter correction (MSC) and mean-centering were applied. It was possible to find clusters representing vitreous and floury endosperm (different types of endosperm present in varying ratios in hard and soft kernels) as well as a third type of endosperm by interactively delineating polygon based clusters in the score plot of the second and fourth principal components and projecting the results on the image space. Chemical interpretation of the loading line plots shows the effect of starch density and the protein matrix. The vitreous and floury endosperm clusters were used to make a partial least-squares discriminant analysis (PLS-DA) model, using four components, with a coefficient of determination (R2) for the y data (kernel hardness category) for the training set of over 85%. This PLS-DA model could be used for prediction in a test set. We show how the prediction images can be interpreted, thus confirming the validity of the PCA classification. The technique presented here is very powerful for laboratory studies of small cereal samples in order to produce localized information. © 2009 American Chemical Society.
dc.description.versionArticle
dc.identifier.citationJournal of Agricultural and Food Chemistry
dc.identifier.citation57
dc.identifier.citation19
dc.identifier.issn218561
dc.identifier.other10.1021/jf9018323
dc.identifier.urihttp://hdl.handle.net/10019.1/10067
dc.subjectstarch
dc.subjectvegetable protein
dc.subjectarticle
dc.subjectchemistry
dc.subjectdiscriminant analysis
dc.subjectmaize
dc.subjectnear infrared spectroscopy
dc.subjectplant seed
dc.subjectregression analysis
dc.subjectDiscriminant Analysis
dc.subjectLeast-Squares Analysis
dc.subjectPlant Proteins
dc.subjectSeeds
dc.subjectSpectroscopy, Near-Infrared
dc.subjectStarch
dc.subjectZea mays
dc.subjectZea
dc.subjectZea mays
dc.titleNear infrared hyperspectral imaging for the evaluation of endosperm texture in whole yellow maize (Zea maize L.) Kernels
dc.typeArticle
Files
Collections
Stellenbosch University - Scopus Publications