The Detection Method Of Fabric Defects

With the development of the integration of global resources,more effective methods of quality testing requirements of unprecedented grow rapidly. Fabric as a necessity in daily life, its quality directly affects the quality of our life. However, as people pursue the high quality of life, the practicality and durability fabric have been unable to meet the needs of the people. They care more about fashion and beauty. It's the key to preventing defects in fabric. Especially for patterned fabric, the defects seriously affect them. By fabric defect detection, we can find the defects that exist in fabric timely and take appropriate measures to effectively reduce the number of flaws in fabric, so the detection of defects for improving the fabric textile quality is essential.In order to improve the efficiency of fabric defect detection, fabric defect detection methods are more and more concerned. As a result, automatic detection system came into being. Automatic detection system not only improves the speed and the accuracy of fabric defect detection, but also reduces the cost of manual inspection. As the entire automatic detection system for fabric involves a lot of knowledge, this article only focuses on the textile defect detection algorithm.The first chapter introduces the meaning of the automatic detection system for fabric, and the commonly used in textile defect detection nowadays. The most commonly used textile defect detection algorithms are mainly divided into three categories: Statistical Approaches, Spectral Approaches, Model-based Approaches. We use statistical approaches to find defects in textile images by defective parts, which is by finding unique statistical characteristics in textiles. There are key assumptions in this process: the statistical characteristics of the image that has no defects is fixed. Uniform texture image is some form of basic texture primitives and displacement with repetitive deterministic rules. Texture primitive with high cyclical, such as textile yarn, can be detected by checking the spectral characteristics of defects. Real texture can be modeled as stochastic process, and texture image or images can be viewed as real space parameters of the probability distribution of the sample observation. The advantage of this model is that it can produce texture that can be observed to match the texture. Defect detection can be considered as derived from the model of statistical hypothesis testing statistics.Chapters II and III, respectively, describe an improved method for LBP textile to defect detection and multi-angle Gabor filters for textile flaw detection. Local Binary Patterns(LBP), has been used earlier on grayscale and rotation invariant classification of Arts and Sciences. Flaw detection in textiles, which is the original LBP method ,is very much due to values.It is not easy for the statistics. And the same value contains too little pixels, which would not achieve the role of statistical information. It is the flaws in textiles Detection, so an improved method of LBP is used. Improved method of LBP is with low complexity and well behaved. Gabor transform is a kind of Windowed Fourier Filter Method, and Gabor function can be of different scales in the frequency domain, different directions extract relevant features. Gabor function and other biological effects are similar to the human eye, so it is often used as a texture recognition, and have good results. Gabor multi-angle multi-scale algorithm is to do four image pyramid image, and then to do at each level of the Gabor filters in four directions, the results of the final fused image obtains a graph.Chapter IV of this article improved the algorithm described above. It is the combination of the improved method and multidimensional LBP angles of Gabor filter that makes a faster and better result.