Using a simple digital camera and SPA-LDA modeling to screen teas

PAULO D. DINIZ; HEBERTTY V. DANTAS; KARLA T. MELO; MAYARA F. BARBOSA; DAVID P. HARDING; ELAINE L. NASCIMENTO; MARCELO F. PISTONESI; FERNÁNDEZ BAND BEATRIZ; MARIO UGULINO DE ARAÚJO

Analytical Methods

Royal Society of Chemistry

Lugar: Londres; Año: 2012 vol. 4 p. 2648 - 2652

1759-9660

Classification or screening analysis of natural unprocessed teasusing simple digital images and a variable selection algorithm isdescribed. The proposed methodology uses color histograms generatedon free downloadable software ImageJ 1.44p as a source ofanalytical information. Two chemometric methods were comparedfor classification of the resulting images, namely Soft IndependentModeling of Class Analogy (SIMCA), and Linear DiscriminantAnalysis (LDA) with variable selection by the Successive ProjectionsAlgorithm (SPA). The results were evaluated in terms of errorsfound in a sample set separate from the modeling process. Thechoice of more informative photometric color attributes (red-greenblue(RGB), hue (H), saturation (S), brightness (B), and grayscale)for screening the tea samples was made during the color modelingbecause SIMCA failed to give good results. Therefore the datatreatment used SPA-LDA, which correctly classified all samplesaccording to their geographical regions, whether from Brazilian,Argentinian or foreign soils.