Abstract: The real-time system including the image acquisition, signal processing and illumination device was developed to detect the weft knitted fabric defects in this paper. The cellular neural network was applied for detecting the defects with obvious gray level distribution and irregular shape. While the line detection based on Radon transform was first proposed to indentify the defects according to their linear shape, because the gray level distribution of the image and background of such defects are very similar. The experiments indicated that the algorithms could effectively detect defects like hole, dropped stitch, fly, float, course mark and miss tuck.