A method is needed to accurately and rapidly determine the gravimetric
bark content of a cotton sample. Gravimetric bark content represents
the percent bark mass throughout the volume of a cotton sample. The cu
rrent method for measuring gravimetric bark content is a labor intensi
ve, lengthy process. Machine vision, on the other hand, is a fast, ine
xpensive method to measure this bulk cotton property. Ten acquired ima
ges of surfaces throughout each sample are used. Classical digital ima
ge processing techniques isolate foreign matter regions in monochrome
video images. Geometric properties (area and perimeter) are used to id
entify which foreign matter is bark and to predict the gravimetric bar
k content in forty-eight cotton samples with varying bark and total fo
reign matter content. We suggest a model with six features and interce
pt, which has an estimated error of 0.46% bark mass.