Application and Development of Digital Image processing Technology in Textile testing

电子科技大学 格拉斯哥学院 2017级 屈颖珊 同组同学 刘奕昕

In recent years, with the development of science and technology, all kinds of advanced technology in textile industry, the application of digital image processing technology in the textile industry is changing, constantly develop its quick, accurate, stable and simple advantages, to a large extent accelerates the pace of textile testing, improves the level of textile testing at the same time.

1 overview of digital image processing technology

Digital Image Processing, also known as computer image processing, refers to the process of converting image signals into digital signals and using computers for processing. It mainly includes the following aspects: digital image acquisition and digitalization, image compression and coding, image enhancement and recovery, image segmentation, image analysis, etc.

In practical application, there are many systems using image processing technology. The general process is: information acquisition - preprocessing - feature extraction - image analysis. Image method is varied, can be directly, or learn through light microscope and electron microscope methods such as access to images taken by enlarged, and then through the A/D conversion, digital image signal, and then the data image processing system ", use computer powerful data processing ability, analysis of images, the output according to the requirement of various indicators. At present, many methods have gradually moved from the stage of theoretical and methodological exploration and research to industrial practical production and application, such as wavelet transform, neural network, expert system, stereo vision, etc., and intellectual energy analysis has become an inevitable trend of research.

Application of digital image processing technology

In the mid-1990s, the application of image processing technology in textile research focuses on: fiber material performance testing, yarn performance analysis, semi-product quality testing and so on. In recent years, people's research focus mainly focuses on the analysis of fabric surface characteristics, automatic analysis of tissue structure, finished and semi-finished product performance testing, etc., some of these technologies have been applied in textile production. In addition, the evaluation of nonwoven fiber and fiber orientation, fiber and yarn performance analysis and other aspects of the research is also deepening.

2. Identification of cashmere and wool fibers

The identification of cashmere and wool has always been a difficult problem in the textile industry. In the past, after the fiber is tested by optical or electron microscope, it still needs the experience of testers to determine that there is no objective and unified evaluation standard, and the use of digital image processing technology is an effective method to solve this problem.

Information science center of Peking University state key laboratory of visual and auditory information processing cashmere wool image automatic identification scheme was proposed, this method is to use automatic threshold method for image binarization, then more straight line cutting, and extract the cashmere wool fineness feature, then the Canny operator was used to extract the edge, on edge of the chart again based on the characteristics of cashmere wool scales image cashmere wool scales length feature extracting, the recognition by the Bayes method into the line. The results show that the method is rapid and accurate in detecting cashmere wool.

F. H. She, I... X Kong et al. studied the identification of merino fine wool and mohair with image processing technology and artificial neural network technology, and developed an intelligent system that can automatically identify the two fibers by combining image processing and neural network technology.

3. Test evenness and cleanliness of raw silk

There are many raw silk detection items, and the detection indexes that determine the grade of raw silk include: denier deviation, maximum denier deviation, cleanliness, cleanliness, evenness, etc. Among them, evenness and cleanliness detection are important indicators to characterize the grade of raw silk. For a long time, people have been using the blackboard detection method to evaluate the grade of raw silk. This method mainly relies on the judgment of people's sensory eyes, and eye transmission is prone to deviation. Subtle differences may lead to imperceptibles and recognition, and the source of measurement value cannot be traced, making the manufacturing process of test samples long. The adoption of digital image processing technology can replace and expand human visual function, which is conducive to solving a series of problems encountered in the use of traditional methods of measurement.

Automatic cleaning measurement is to classify and count the cleaning according to the standard, calculate the average difference and relative difference between the length, width, circumference, area, center gray and surrounding gray according to the outlined rough defect shape. Doug gang et al. have used the method of feature analysis to judge the types of defects and count the number of them. They have made a video recording, quantitative analysis and feature analysis of a dozen samples of cleaning inspection.

Evenness test is, in fact, the nature of gray level difference, in the image of grey value sampling in two-dimensional gray function, value, to quantify, get with the discretization of the image gray scale two dimensional array said, can be used to calculate machine for all kinds of processing, transform the image into a measurement of the image, then USES the means of pattern recognition, discriminant is made of raw silk evenness. Hu Zhenzhou etc. Using the method of micro optical imaging measurement of raw silk evenness, its concrete process strips is driven by speed controlled traction device, from a microscopic lens by, strips image amplified by microscope, image on the CCD image sensing elements, the CCD image sensing elements to video signal output signal acquisition card, after signal acquisition card processing, become available for identification and calculation of computer digital signal, after further processing, testing data are obtained.

4. Test yarn blending ratio

The blending ratio of blended yarn is generally measured by chemical analysis and microscope cross section observation. Due to the rapid renewal rate of chemical fibers, chemical analysis and test methods also need to be constantly updated, which makes it difficult to find a suitable chemical analysis method, increasing the test cost, and the results are not satisfactory. It is a new method to detect yarn blending ratio quickly and accurately by using computer to extract characteristic parameters and automatically identify fibers and measure yarn blending ratio. The morphological parameters of blended yarn are examined by computer, including fiber root count, root count ratio, area ratio, area ratio, primary moment and transfer index. At present, there are many related studies on cotton/linen, silk/wool, linen/polyester, wool/polyester blended yarn, and the research results are relatively mature.

5. Yarn evenness test

Digital image processing technology is used to detect the evenness of yarn, which can not only objectively and accurately assess the yarn grade, but also the detection process is not affected by the environment. In addition, this method can accurately simulate the final cloth surface of the yarn. At present, the methods of measuring yarn evenness by digital image technology are as follows: image yarn dryness meter, image morphology method, image texture analysis method, computer vision technology, etc.

Image type test principle of the yarn evenness tester is to test yarn irregularity of instrument Ⅲ, from a capacitive sensor test to the photoelectric method is adopted to improve the yarn irregularity test; Image morphology method refers to the collection of yarn blackboard image through image preprocessing, image banalization, autocorrelation method, mathematical morphology of the processing of a can be rated image, using this method to test the uniformity of yarn, yarn appearance quality can be objectively assessed; Image texture analysis and yarn evenness test mainly USES image processing and pattern recognition technology to automatically analyze and judge the evenness of blackboard image, so that the grading of blackboard can be automatically implemented. Computer vision analysis system refers to the combination of optics and computer image processing technology, the use of wavelet analysis, Fourier analysis and other application of mathematical knowledge, the scanner, CCD camera and other real-time collection of yarn image image processing, finally get the intuitive yarn fineness irregularity.

6. Fabric surface pilling test

Pilling is one of the most important factors affecting fabric wear performance. It is often caused by wear and tear on the fabric surface. In related tests, the accuracy of traditional measurement methods is not high enough and there is a large proportion of subjective factors involved. However, digital image processing technology overcomes the above problems and can test the fabric surface objectively and accurately.

Chen xia, donghua university, used an image analysis system for objective assessment of fabric pilling grade, first get reflection fabric 2 d contour data from the image, through a set of matching filter testing hair bulb, using the method of block threshold for segmenting the bulb, select number of hair bulb, the bulb area and the bulb volume level evaluation equation, and USES the fuzzy logic system as a grade evaluation model of the final.

Wang et al., using statistical histogram technology to pilling image filtering, sharpening and segmentation preprocessing, and then based on autocorrelation function to determine areas for fabric pilling is not a repeat of the structure shape and size, and application of expansion and corrosion technology dealing with binary image, the original image processing into no organizational structure and only images of the hair bulb, finally determine the size, number and form of the bulb.

Many foreign experts have conducted in-depth studies on this. Konda et al. proposed an objective evaluation method for pilling performance. They processed the image threshold and then compared it with the standard image to establish the level of the number of small balls. Abril et al. also used image analysis to measure the area of small spheres in a standard fabric image.

7. Fabric defect test

After the application of digital image processing technology to fabric defect testing, the testing results have been greatly improved. The key of its test method is how to segment and recognize the right and reasonable texture. The defect detection and measurement system proposed by han wupeng et al. preprocessed the collected images, extracted the features with wavelet transform, and used fuzzy technology for reasoning and identification, which could improve the enhancement effect of the edge of the defect and characterize the feature information of the local defect. Yixiang, FrankZhang and Randal1 of tenaxi university. R. Bresee proposed to use both gray scale statistical method and morphological method to extract feature regions, analyze feature regions, and use pattern recognition to classify defects. Hu Yan [8] put forward a kind of fabric defect based on wavelet transform and morphological edge detection method, using morphology for defect detection, after wavelet decomposition, and then by using the wavelet modulus maxima method and algorithm based on mathematical morphology to extract defect edge, high, low and adopt the reasonable fusion rules which fuse the two edge images, end up with clear and accurate edge.

Another hot issue on fabric defect testing is how to dynamically identify and classify fabric defects. Abroad many years ago, have been engaged in the research and prove this is a feasible method, such as Switzerland Uster company Fatrriscan automatic cloth inspection system, equipped with 2 ~ 8 in the width direction only special high-resolution line scan CCD camera, USES the neural network technology, testing fabric width range is 110 ~ 440 cm, speed up to 120 m/min. Although the research on this aspect has made some progress in China, it has not yet formed a perfect system.

8. Chroma test

At present, there are few researches on the application of image processing technology in textile colorimetry testing. Chromaticity does not include brightness, it is a characteristic of color, it reflects the hue and saturation of color, people usually make visual and physical measurement of color based on the chromaticity system stipulated by CIE. Using image processing technology, R, G and B values can be directly obtained according to the principle of primary colors. The resolution of R, G and B values is 256, which is far beyond the recognition ability of human eyes and can accurately describe the color value of objects, and then calculate the three stimulus values and yellowness, and get the final result. realized the reliability measurement of cotton fiber chromaticity through digital image processing technology according to the chromaticity principle. Compared with traditional measurement, it not only reduced the cost, but also greatly improved the speed and stability of measurement.

9. Testing nonwoven fabrics

Nonwoven fabric is developing rapidly, the product quality is getting higher and higher, and it occupies a larger and larger share in the textile market. Therefore, the detection technology of nonwoven fabric should meet the requirements of quality and develop in a simple, fast, objective and reliable direction, while digital image processing technology makes it possible. At present, digital image processing technology is mainly used in testing nonwoven fabrics, such as fiber orientation test, pore size distribution test, flaw test, uniformity test and fiber diameter test.

10. conclusion

The application of digital image processing technology in textile industry has achieved considerable results at present, but due to the limitation of hardware and software conditions, it has not been fully put into use in many aspects, failing to meet the needs of various aspects of textile testing. The technology will be developed in the direction of intelligence, artificial intelligence, genetic algorithm, fuzzy theory and so on have been applied in textile quality assessment, and two-dimensional image processing has been gradually developed in the direction of three-dimensional. With the continuous development of computer science technology and mathematical theory, the application of digital image processing technology in the textile field will continue to expand and play a greater role. It is believed that the acquisition of new theories and methods through continuous scientific research will promote the development of China's textile industry towards the field of high technology.