The existing machine vision systems acquired the images of the scoured wool with contaminants in the visible lights band by CCD cameras, and then processed the images for the removal of the contaminants from wool. They cannot distinguish the scoured wool from the white contaminants, reducing the quality of the fiber products significantly. In order to overcome this problem, a spectrophotometer was used to measure the diffuse reflection spectrum of scoured wool with nine types of white contaminants in 200~2 600 nm band, the relationship of their diffuse reflection spectrum with the changes of their wavelength was analyzed, and then by constructing an optimal model for band selection, the best band for discriminating white contaminants from wool was calculated. Based on that, the infrared images of scoured wool with the white contaminants were acquired by an infrared camera and analyzed. The experiment indicated that the histograms of the infrared images are bimodal so that their images would be segmented easily. Therefore, this method can be used for detecting write contaminants in scoured wool in a machine vision system.
