Intelligent Recognition Of Silk Fabric Defects

This paper mainly deal with the intelligent recognition of familiar defects such as loom bars, broken filling, broken picks, double filling and stains of tabby, twill, satin fabrics. First SONY digital camera is used to acquire defect images by placing a black background under sample fabric, and used series of pre-image processing method. For example histogram transform is used to increase image contrast, a threshold, which is obtained by calculation, is used for thresholding defect image. Filtering method is used for eliminating single noises that is formed after thresholding process. Thus defect part is being parted from fabric texture, and obtained a fabric defect image for analyzing. Grey level statistical method is used for analyzing fabric defect images which is obtained by pre- processing, and some basic information of each defect image are acquired. Fabric defects intelligent recognition is done by BP neural network. First BP neural network is being trained, then the basic fabric defect character information, which is acquired by gray level statistical processing, is inputted to BP neural network to classify fabric defects.