Thesis (PhD (Food Sc))--Stellenbosch University, 2015. / ENGLISH ABSTRACT: Maize kernels consists of two types of endosperm, a harder vitreous endosperm and a softer floury
endosperm, and the ratio of the vitreous and floury endosperm present mainly determines the
hardness of the kernel. Maize (Zea mays L.) is a staple food in many countries, including South
Africa, and is industrially processed into maize meal using dry-milling. For optimal yield and higher
quality products, hard kernels are favoured by the milling industry. Despite many maize hardness
methods available, a standardised method is still lacking, furthermore, no dedicated maize milling
quality method exists.
Using an industrial guideline (chop percentage), a sample set of different maize hybrids was
ranked based on milling performance. Unsupervised inspection (using principal component
analysis (PCA) and Spearman’s rank correlation coefficients) identified seven conventional
methods (hectoliter mass (HLM), hundred kernel mass (HKM), protein content, particle size index
(PSI c/f), percentage vitreous endosperm (%VE) as determined using near infrared (NIR)
hyperspectral imaging (HSI) and NIR absorbance at 2230 nm (NIR @ 2230 nm)) as being
important descriptors of maize milling quality. Additionally, Rapid Visco Analyser (RVA) viscograms
were used for building prediction models, using locally weighted partial least squares (LW-PLS).
Hardness properties were predicted in the same order or better than the laboratory error of the
reference method, irrespective of RVA profile being used.
Classification of hard and soft maize hybrids was achieved, based on density measurements
as determined using an X-ray micro-computed tomography (µCT) density calibration constructed
from polymers with known densities. Receiver operating classification (ROC) curve threshold
values of 1.48 g.cm-3
, 1.67 g.cm-3 and 1.30 g.cm-3 were determined for the entire kernel (EKD),
vitreous (VED) and floury endosperm densities (FED), respectively at a maximum of 100%
sensitivity and specificity.
Classification based on milling quality of maize hybrids, using X-ray µCT derived density and
volume measurements obtained from low resolution (80 µm) µCT scans, were achieved with good
classification accuracies. For EKD and vitreous-to-floury endosperm ratio (V:F) measurements,
93% and 92% accurate classifications were respectively obtained, using ROC curve. Furthermore,
it was established that milling quality could not be described without the inclusion of density
measurements (using PCA and Spearman’s rank correlation coefficients).
X-ray µCT derived density measurements (EKD) were used as reference values to build NIR
spectroscopy prediction models. NIR spectra were acquired using a miniature NIR
spectrophotometer, i.e. a microNIR with a wavelength range of 908 – 1680 nm. Prediction statistics
for EKD for the larger sample set (where each kernel was scanned both germ-up and germ-down)
was: R2
V = 0.60, RMSEP = 0.03 g.cm-3
, RPD = 1.67 and for the smaller sample set (where each
kernel was scanned only germ-down): R2
V = 0.32, RMSEP = 0.03 g.cm-3
, RPD = 1.67. The results from the larger sample set indicated that reasonable predictions can be made at the fast NIR scan
rate that would be suitable for breeders as a rough screening method. / AFRIKAANSE OPSOMMING: Mieliepitte bestaan uit twee tipes endosperm, ‘n harder glasagtige endosperm en ‘n sagter
melerige endosperm, en die verhouding waarin die twee tipes endosperm aangetref word, bepaal
hoofsaaklik die hardheid van die pit. Mielies (Zea mays L.) is ‘n stapelvoedsel in baie lande,
insluitende Suid-Afrika, en word industrieël geprosesseer na mieliemeel deur van droë-vermaling
gebruik te maak. Vir optimale produksie en beter kwaliteit produkte, word harde pitte deur die
meule verkies. Ongeag die beskikbaarheid van verskeie mielie hardheid metodes, ontbreek ‘n
gestandardiseerde metode nog, en verder bestaan ‘n metode om mielies se maalprestasie te
bepaal ook nie.
‘n Monsterstel, bestaande uit verskillende mieliebasters, is op grond van maalprestasie
ingedeel deur van ‘n industriële riglyn (chop persentasie) gebruik te maak. Inspeksie sonder toesig
(deur gebruik te maak van hoofkomponentanalise (HKA) en Spearman’s
rangkorrelasiekoëffisiënte) het sewe onkonvensionele metodes (hektoliter massa, honderd pit
massa, protein inhoud, partikel grootte indeks, persentasie glasagtige endosperm soos bepaal
deur gebruik te maak van naby-infrarooi (NIR) hiperspektrale beelding en NIR absorbansie by
2230 nm) identifiseer as belangrike beskrywers van maalprestasie. Daarbenewens, is Rapid Visco
Analyser (RVA) viskogramme gebruik om voorspellingsmodelle te bou deur gebruik te maak van
plaaslik geweegte gedeeltelike kleinstekwadrate (PG-GKK) wat hardheidseienskappe kon voorspel
met laer, of in dieselfde orde, laboratorium foute van die verwysingsmetodes, ongeag die gebruik
van verskillende RVA profiele.
Klassifikasie tussen harde en sagte mieliebasters was moontlik, gebasseer op
digtheidsmetings soos bepaal met ‘n X-staal mikro-berekende tomografie (µBT) digtheids
kalibrasie gebou vanaf polimere met bekende digthede. Ontvanger bedryf kenmerkende (OBK)
kurwe drempelwaardes van 1.48 g.cm-3
, 1.67 g.cm-3 en 1.30 g.cm-3
is bepaal vir hele pit, glasagtige
en melerige endosperm digthede, onderskeidelik, teen ‘n maksimum van 100% sensitiwiteit en
spesifisiteit.
Klassifikasie van die mieliebasters, gebasseer op maalprestasie en deur gebruik te maak van
X-straal µBT afgeleide digtheid en volume metings soos verkry teen lae resolusie (80 µm)
skanderings, was moontlik met goeie klassifikasie akkuraatheid. Vir heel pit digtheid en glasagtigtot-melerige
endosperm verhouding metings is 93% en 92% akkurate klassifikasies verkry
wanneer OBK kurwes gebruik is. Verder is dit vasgestel (deur gebruik te maak van HKA en
Spearman’s rangkorrelasiekoëffisiënte) dat digtheidsmetings ingesluit moet word vir ‘n volledige
beskrywing van maalprestasie.
X-straal µBT afgeleide digtheid metings is gebruik as verwysings waardes om NIR
spektroskopie voorspellings modelle te bou. NIR spektra is verkry deur van ‘n miniatuur NIR
spektrofotometer, naamlik ‘n microNIR, bebruik te maak vanaf 908 – 1680 nm. Voorspellings
statestiek vir die groter monsterstel (waar elke pit beide kiem-bo en kiem-onder geskandeer is) was vir HPD: R2
V = 0.60, RMSEP = 0.03 g.cm-3
, RPD = 1.67 en vir die kleiner monsterstel (waar
elke pit was slegs kiem-onder geskandeer is) vir HPD: R2
V = 0.32, RMSEP = 0.03 g.cm-3
, RPD =
1.67. Die resultate van die groter monsterstel het aangedui dat redelike voorspellings moontlik is,
teen die vinnige NIR skaderings tempo wat as rowwe vertoningsmetode geskik sal wees vir telers.
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:sun/oai:scholar.sun.ac.za:10019.1/96967 |
| Date | 04 1900 |
| Creators | Guelpa, Anina |
| Contributors | Manley, Marena, Geladi, Paul, Du Plessis, Anton, Stellenbosch University. Faculty of Agrisciences. Dept. of Food Science. |
| Publisher | Stellenbosch : Stellenbosch University |
| Source Sets | South African National ETD Portal |
| Language | en_ZA |
| Detected Language | Unknown |
| Type | Thesis |
| Format | xvii, 119 pages : colour illustrations |
| Rights | Stellenbosch University |
Page generated in 0.0023 seconds