Objective Evaluation of Soil Release in Fabrics

Abstract

Soil release is an important attribute of fabrics that impacts their pricing on the marketplace. Currently, stain release is evaluated manually and hence its assessment tends to be subjective. This paper introduces a novel approach based on image analysis for objective evaluation of stain release in fabrics. This system has the capability to accurately detect, localize and grade stains with minimal human intervention. The grades assigned by the system are based on a digitized version of the AATCC soil release replica. The stain detection performance of the system was validated on a large dataset of 360 stain images. The system produced excellent results with a sensitivity (true positive rate) of around 93 percent and specificity (1 — false positive rate) of around 95 percent. Further, a correlation study and analysis of variance were performed using the system assigned grades and the technician assigned grades, and these initial results are very promising.