All publications from your selection

Download RAW file:
DOI:
Description: Hyperspectral imaging (HSI) has shown great potential for rapid and nondestructive assessment of food quality and safety. Applications of the technique result in large data sizes with significant reductant variables. There are also multicollinearity, particle size variation and light scattering effects that present difficulty in predictive modelling. The goal of this study was to investigate the effectiveness of various preprocessing techniques that could be used to mitigate some of the challenges associated with HSI data. Different preprocessing techniques namely: PCA, SNV, OSC, MSC, log-normalization, 1st and 2nd derivatives, detrend, mean centering, and autoscaling were studied. HSI datacube obtained from chicken egg was used in the study and the ZeroR algorithm was implemented to estimate baseline performance for an eventual subsequent modelling. The results show that autoscaling, log-normalisation and OSC methods performed optimally in terms of observed classification accuracies, root mean square errors and percentage of information explained by various principal components. The said preprocessing techniques were then recommended for use in handling chicken egg hyperspectral data prior to a downstream multivariate analysis.

Keywords: hyperspectral data, chicken egg, preprocessing techniques, ZeroR algorithm.
Résumé:
Mots-clés:
Citation:
Volume:
Issue:
Pages -
Contributor:
Date: 2017-08-07
Technical field:
Conference name: CSBE/SCGAB 2017 Annual Conference, Canad Inns Polo Park, Winnipeg, Manitoba, 6-10 August 2017.
Session name: Session 6C: Food Processing: Thermal and Non-thermal Approaches

Other information:
Type: Text.Article
Format: PDF
Publication type: Technical conference
Source:
Relation:
Coverage: Canada
Language 1: en
Language 2:
Rights: Canadian Society for Bioengineering
Notes:
Other files: