Research On Automatic Detection Of Fabric Defects Based On Machine Vision

Nowadays, quality control is a crucial link in the textile industry, besides, the fabric defect detection is the most important part of the quality control links. Due to the shortcoming of manual detection, such as low efficiency, high missing rate, affection by emotion and so on. So it is an inevitable trend that automatic system instead of manual detection in fabric defect field. Currently, most of the automatic equipment in our country is imported, these instruments cost much and the adaptability is poor. Therefore, it is necessary for china to have its own automatic inspection machine.This paper is on the basis of theoretical research in related fields in domestic and overseas, the technology of machine vision and images collection of textile are discussed and researched. Currently, the obstacle of automatic inspection system is images processing are slow, it is difficult to guarantee the real-time detect. If we use the same algorithm to detect most of the fabric, the complex fabric defect detection algorithm must be used. But to simple fabric defects, these algorithms are excessive, for some of image processing is unnecessary. So in this project, after building the automatic acquisition system, giving different algorithms according to different fabric defects included weft, catkins, knot, and broken, broken hole. In this way, the automatic system will use simple algorithm to detect simple fabric, which improves the speed of image processing, provides the possibility for real- time detection of fabric defects.The main work of this project is as follows. First, build automatic system of fabric image collection system. The system includes light source, image sensor, CCD camera, optical lens and software platform. Second, introduce the image processing method in this experiment, which includes the gray image, the filtering of image, the image segmentation, the edge extraction and the extraction of image feature. Last, give different algorithms according to different fabric defects, and using the proposed algorithms for different fabric defects are validated. The experimental results are given in the end.