95%。 For the existence of foreign fibre during cotton processing, machine vision technology and image processing were used in order to not only extract foreign fibre goals, but also collect characteristic data of foreign fibre. Decision tree theory and feature vectors extracted were used to recognize foreign fibre after eigenvectors of aimed foreign images were effectively extracted through an improved version of rough set theory. Experimental results showed that feature vectors extracted from the image of foreign fibre for the identification of cotton foreign fibre was effective and the recognition rate reached more than 95%.