Recognition Algorithm Research Of Fabric Defect On-line Detection

With the development of industrial automation, textile industry automation is inevitable. With more and more people pay attention to the quality of products, improving the quality of products and accelerating the auto-detection technology is a priority. This article is based on machine vision technology to study about recognition algorithm of fabric defect on-line detection, completed study of fabric defect on-line detection and implementation of key technologies.First, by creating the classification standards of fabric defect and the characteristics of fabric defects analysis, it offers gist for the identification of classification algorithms, fabric On-line defect detection system project, discusses the principles of fabric defect identification, according to actual requirements, from hardware and software analysis of the process of system implementation are discussed, focusing on image processing.Then, presented to the gray image pre-processing, it is to make the image equalization, the median filter and dislocation difference algorithm, the image is removed noise and reduced the shadow texture of fabric for the feature extraction to provide higher quality images. The image feature extraction is proposed for fabric defect on-line rapid identification and classification. The image of the fabric structure in the spatial domain on the straight square-wave and wavelet transform, extract the image feature value. They are used to distinguish defects. If defect image is the defect, it is obtained binary images, using mathematical morphology to extract the geometric parameters of fabric defects, for fabric defect classification. In this paper, quickly and conveniently analyze algorithms by Lab VIEW software, and program by Visual C++, combined with computer-based instruction set hardware-optimized IPP library for feature extraction algorithm. Finally, for the fabric defect identification and classification, bring two neural network classification methods. One neural network is used to distinguish image whether there are defects, another one is used to defect classification. Based on the standard BP network algorithm, brought an improved BP algorithm, and applied to on-line detection. The results show that the training speed and recognition classification accuracy improve by the improved BP algorithm...