Biscuit quality assessment by spectral imaging | Jens Michael Carstensen
| | Abstract | A fast, non-destructive, versatile method for assessing biscuit quality is presented. The method
integrates color (or browning) and humidity assessment with dimensional measurements. In the
quality control function of biscuit production this information is important to optimize production
processes and to assure quality of the final product.
Spectral imaging and analysis of one biscuit is done in a few seconds. The imaging system,
VideometerLab (www.videometer.com) captures 18 high resolution spectral images in the range
400-1000 nm. Acquisition is band-sequential and performed by LED strobing into an integrating
sphere illumination geometry. The spectral image will contain a NIST calibrated reflectance
spectrum in every pixel.
Biscuit color is estimated either in CIELAB coordinates, or as a browning index. Both
representations may be estimated in both the biscuit center region and in the edge/corner region.
CIELAB coordinates are spectrally reconstructed from the reflectance spectra, and the baking
index is constructed using a normalized canonical discriminant analysis.
A biscuit humidity model has been built on regional spectral imaging features through stepwise
feature selection and linear regression towards conventional weight humidity measurements. The
data set includes biscuits from different lines and both hot and cooled biscuits. Within the relevant
humidity range 1%-3% we obtain a RMSEP of 0.14%.
Dimensional measurements relates to the packaging of the biscuits and consist of height and
lateral dimensions, e.g. length and width, and height. | | Type | Conference paper [With referee] | | Conference | Proceedings of the Third Annual Hyperspectral Imaging Conference | | Year | 2012 Month May | | Publisher | Istituto Nazionale di Geofisica e Vulcanologia | | Address | Roma, Italy | | Electronic version(s) | [pdf] | | BibTeX data | [bibtex] | | IMM Group(s) | Image Analysis & Computer Graphics |
|