Abstract:2倍多,。In computer-vision-based systems for detecting foreign fibers, an efficient segmentation method is vital for improving the speed and precision of detection. An image enhancement model was presented based on histogram analysis, an improved Otsu's method was proposed for segmenting gray images of foreign fibers, and the segmentation results were analyzed in detail. The results indicate that all histograms of gray images of foreign fibers are of single peak, and the contrast between object and background can be remarkably enhanced by the enhancement model presented. The Otsu's method is effective for images of sheet foreign fibers, but does not work well for images of wirelike or villiform fibers, as some background pixels with gray values closed to the threshold are easily segmented as object pixels by mistake. But after the images were enhanced, the satisfactory segmentation results were achieved. The improved Otsu's method reduced the searching range for calculating optimal threshold from 0~255 to 150~230. The calculating speed in this stage was improved more than twice.