Automatic Structure Recognition Of Woven Textile

Researches of automatic recognition of woven textile were motivated by the demands of fabric industry. Humans have a surprising capability to easily comprehend an image, but studies of automatic recognition of woven textile have always been difficult mainly because the physical irregularity of fibers and textures under illumination, and also lack of prior knowledge in recognition.The principal purpose of this dissertation is to find out some practical methods for industrial recognition for some certain types of woven texture with attempts. Some general prerequisites are required.Some techniques are proposed in this dissertation based on the integration of certain fundamental image process methods. The numerical experiments of case study in this dissertation support the methods. The work described here has not yet provided a completely solution, but it suggests that the use of these methods combined with the human manipulations interactively makes it possible to recognition of woven textures if the textures are not too complex.Experimental results show that this technique produces satisfied results than traditional techniques based on unitary specific method. And industrial application will be expected in a not-too-long future.Future work will be focused on the development of industrial software.