Masters Theses

Date of Award

8-1998

Degree Type

Thesis

Degree Name

Master of Science

Major

Textiles, Retail, and Consumer Sciences

Major Professor

Kermit Duckett

Committee Members

Randall R. Bresee, Gajanan Bhat

Abstract

Cotton color grading has always represented one of the main factors of cotton price determination. First cotton color evaluation was administered by visual inspection of the samples by a trained and experienced person - the classer. Since 1929, there has been great effort to construct a color measuring device to replace the human factor. Even though the cotton colorimeter has undergone a great deal of improvement and sophistication, there is just 70% agreement between grades assigned to the sample by the classer and by the colorimeter. Today, the cotton colorimeter is part of the High Volume Instrument system (HVI) - an instrumentation complex designed to measure the most important properties of cotton fibers - and is used only as assisting equipment which complements the classer’s grading. The objective of this work is to compare the current cotton color measurement system with spectrometer and image analysis techniques in order to minimize the disagreement in classer and instrument grading. It was shown that there are several limiting features in current cotton color grading. Currently, cotton color measurements are represented by two color parameters - reflectance (Rd) and yellowness (+b), neglecting the other of the three color parameters. Also, color measurements obtain an average value of the whole sample area, disregarding any color variability within the sample area. By visual inspection, the classer interprets complex information about color, color variability, texture and trash particles within the sample. In order to investigate the influence of other color variables on the accuracy of correct grade determination, spectrometer color measurements and image analysis characteristics were obtained for the two sample sets (N1=12, N2100) provided by USDA. From the first set, it was shown that redness values of certain samples were unusually high and it was proposed that those values may provide additional data that could be used to raise the level of grading agreement between classer and instrument. The second sample set consisted of 50 samples that received identical grading and 50 samples that received different grading between classer and instrument. By applying newly developed software and incorporating statistical methods, it was shown that using three instead of two color parameters would increase the agreement level by 44%. When the two currently used color parameters were supported only by variance of yellowness, the agreement improved by 26%. When yellowness contrast was used with the two HVI parameters, the agreement increased by 9% and when yellow spot area information complemented the HVI data, the agreement increased by 7%. It was also shown that, the level of agreement reached 95% when all the known parameters were together used to determine the grade. Redness alone was the major contribution to agreement between instrument and classer.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS