Quantification of <Emphasis Type="Italic">Lactobacillus</Emphasis> in fermented milk by multivariate image analysis with least-squares support-vector machines |
| |
Authors: | Alessandra Borin Marco Flôres Ferrão Cesar Mello Lívia Cordi Luiz C M Pataca Nelson Durán Ronei J Poppi |
| |
Institution: | (1) Chemistry Institute, Campinas State University, P.O. Box 6154, 13084-971 Campinas, SP, Brazil;(2) Chemistry and Physics Department, Santa Cruz do Sul University, C.P. 188, 96815-900 Santa Cruz do Sul, RS, Brazil;(3) Chemistry Institute, Franca University, C.P. 32, 14404-600 Franca, SP, Brazil;(4) Environmental Science Center, Mogi das Cruzes University, 08780-911 Mogi das Cruzes, SP, Brazil |
| |
Abstract: | This paper reports an approach for quantification of Lactobacillus in fermented milk, grown in a selective medium (MRS agar), by use of digital colour images of Petri plates easily obtained
by use of a flatbed scanner. A one-dimensional data vector was formed to characterize each digital image on the basis of the
frequency-distribution curves of the red (R), green (G), and blue (B) colour values, and quantities derived from them, for
example lightness (L), relative red (RR), relative green (RG), and relative blue (RB). The frequency distributions of hue,
saturation, and intensity (HSI) were also calculated and included in the data vector used to describe each image. Multivariate
non-linear modelling using the least-squares support vector machine (LS-SVM) and a linear model based on PLS regression were
developed to relate the microbiological count and the frequency vector. Feasibly models were developed using the LS-SVM and
errors were below than 10% for Lactobacillus quantification, indicating the proposed approach can be used for automatic counting of colonies. |
| |
Keywords: | Multivariate image analysis Colour Lactobacillus Fermented milk Least-squares support vector machines |
本文献已被 PubMed SpringerLink 等数据库收录! |
|