The type and percentage of fiber contained in textile fabrics are important factors affecting the quality of fabrics, and they are also what consumers pay attention to when buying clothing. The laws, regulations and standardization documents related to textile labels in all countries in the world require almost all textile labels to indicate fiber content information. Therefore, fiber content is an important item in textile testing.
The current laboratory’s determination of fiber content can be divided into physical methods and chemical methods. The fiber microscope cross-sectional measurement method is a commonly used physical method, including three steps: the measurement of fiber cross-sectional area, the measurement of fiber diameter, and the determination of the number of fibers. This method is mainly used for visual recognition through a microscope, and has the characteristics of time-consuming and high labor cost. Aiming at the deficiencies of manual detection methods, artificial intelligence (AI) automated detection technology has emerged.
Basic principles of AI automated detection
(1) Use target detection to detect fiber cross-sections in the target area
(2) Use semantic segmentation to segment a single fiber cross section to generate a mask map
(3) Calculate the cross-sectional area based on the mask map
(4) Calculate the average cross-sectional area of each fiber
The detection of blended products of cotton fiber and various regenerated cellulose fibers is a typical representative of the application of this method. 10 blended fabrics of cotton and viscose fiber and blended fabrics of cotton and modal are selected as the test samples.
Place the prepared cross-section sample on the stage of the AI cross-section automatic tester, adjust the appropriate magnification, and start the program button.
(1) Select a clear and continuous area in the picture of the fiber cross section to draw a rectangular frame.
(2) Set the selected fibers in the clear rectangular frame into the AI model, and then pre-classify each fiber cross section.
(3) After pre-classifying the fibers according to the shape of the fiber cross-section, image processing technology is used to extract the contour of the picture of each fiber cross-section.
(4) Map the fiber outline to the original image to form the final effect image.
(5) Calculate the content of each fiber.
For 10 different samples, the results of the AI cross-section automatic test method are compared with the traditional manual test. The absolute error is small, and the maximum error does not exceed 3%. It conforms to the standard and has an extremely high recognition rate. In addition, in terms of test time, in traditional manual testing, it takes 50 minutes for the inspector to complete the test of a sample, and it only takes 5 minutes to detect a sample by the AI cross-section automatic test method, which greatly improves the detection efficiency and saves manpower and time cost.