Fabric Defect Detection Based On Image Processing Technology

In the process of textile manufacturing, the detection and control of the quality is an important step, while the core process of the step is the capture and repair of textile flaws. Now, through the operation of the workers by virtue of observation to find flaws is the method of most light industrial factories. There are some problems, such as labor intensity, high undetected and false detection rate. Moreover, manual inspection will be affected by the operators’ subjective reasons, thus can not ensure its higher accuracy. The eyes of the operators are also prone to be damaged. Fabric defect detection is a research focus in the field of computer vision, and there have been a lot of research results. Because the current detecting technology can’t fully meet the needs of actual production from the point of effectiveness, real-time performance and applicability, so it is necessary to explore new technology of fabric defect detection. After conducting research on many current algorithms for fabric defect detection, the study of project as follows:(1) Classification methods are discussed in detail, mainly include the defect classification based on the geometric characteristics and texture features. In addition, the hardware and software environment constituting fabric defects detection was introduced.(2) Fabric defect detection methods are summarized, mainly including the methods based on spatial domain and frequency domain. Also the image digitization and its constitute principle, the basics of image preprocessing and image segmentation are introduced.(3) Until now, textile defect automatic detection has still been a difficult problem. In this article, to begin with an important visual feature of texture, a new approach is proposed for fabric defect detection based on texture periodicity analysis according to feature of fabric texture image by learning the textile flaw image and its corresponding flawless image. The textile defect detection process based on the periodic analysis as follows:image preprocessingâ†'operation of the autocorrelation functionâ†'operation of the basic unit templateâ†'enhancement of the defective areaâ†'constructing a mean imageâ†'image segmentation with Otsuâ†'median filterâ†'defect localization.(4) Another approach is proposed for textile defect detection based on Gabor filter masks in this thesis. This is a well-known algorithm, it has a good ability to fabric analysis in the time domain and frequency domain. The artificial vision mode can be simulated by a set of self-similar Gabor filters with different direction and scale characteristics. This topic uses binarization algorithm for image segmentation to complete the textile flaw detection and takes defect detection method using multi-channel Gabor filtering to achieve the purpose of improving multi-channel information fusion algorithm. Experimental data proves that the method is able to select the channels which fit the human visual system to integrate information.In addition, the program design is completed in MatlabR2007b image processing software platform. The program achieves the collection of the textile image with BMP file, defect detection and the results show. It completes the final debugging and testing.