Color discrimination of biological material

Abstract

A statistical classifier is developed to discriminate between biological material having high as well as low color contrast. Color indexing is used to obtain probability distributions of different classes contained in RGB color digital images. Polynomial regression is employed to get the best-fit equation for class distribution. The polynomial is used to extract the desired class by a thresholding technique. Color images of redroot pigweed and soybean with soil as background are processed to evaluate the method effectiveness. Three classes are targeted, namely, weed stem, leaf, and soil. Color indexing is found useful in color-based image classification. Classification results proved that the classifier built using polynomial functions performed effectively with low misclassification.