Multi-focal Image Fusion And Its Application In Textile Digital Testing

Textile testing aims to test the performance of textile materials and to evaluate their quality so as to provide a basis for production and trade.Therefore,it is necessary to continuously improve the textile testing ability and technical level.With the development of the information science,the detection of textile products using digital image processing technology has become the development trend of textile testing,with the advantages of rapid and accurate testing and avoiding the requirement of high-precision equipment.At present,image processing has been widely used in textile testing,such as fiber diameter testing,yarn twist and fineness testing,fabric structure and non-woven orientation testing.When image processing technology is used to detect,except for the requirements of a variety of image processing algorithms,the clear and complete unprocessed image is also the key factor to accomplish the accurate detection.In the optical system,the depth of field is limited.Therefore,when the spatial distribution of the textile exceeds the depth of field distance,the microscopic images of the textile obtained by using an optical microscope tend to have a problem of multi-focal plane,which are defined as multi-focal images.The multi-focal images are sharp within the focusing range and are blurred out of focus range.In the subsequent image processing,fuzzy information will affect the detection accuracy.In order to solve the image multifocal problem and to facilitate the application of image processing technology in textile testing,a new image fusion algorithm-Regional Gradient Variance is proposed in this paper.With the help of image system,a clear and complete fused image of textile is successfully obtained.Based on the fused image and with image feature extraction methods and data analysis,fiber identification and fibers contents,yarn hairiness value and nonwovens straight-through porosity testing are accomplished.Through the test results and comparative experimental anaylsis,the fused image perserves the clear image information and improves the accuracy of textile digital detection.The automatic detection syten constructed in the paper can quickly and accurately realize the textile detection and facilitate the useability of textile testing.The main contents of this thesis can be summarized as follows:(1)Propose a new multi-focal image fusion algorithm-Regitional Gradient Varience Algorithm based on the pixel level spatial method and combine the fusion algorithm into the image acquisition system to acquire instantly the fused imageBased on the depth of field of the microscope system and the spatial state of the textile,multi-layer serial microscopic multi-focal images located in different focus location((5-axis direction)and the same field of view((3-axis and (4-axis directions)are obtained through the movement of the microscope platforms (3,(4 and (5 directions.For the same pixel coordinate position(,),the pixel has serial focus positions defined as image layer numbers.Compared with existing image fusion algorithms,a new multi-focal image fusion algorithm-Regitional Gradient Varience Algorithm is propsed,where a new pixel sharpness evaluation function and focus position consistency check of adjacent pixels are implemented to find the optimal focus position from the serial focus positions.By point to point mapping,the clear image information located in the optimal focus position is extracted to fully integrate a fused image.The fusion algorithm is added to the image acquisition system to realize the immediate acquisition of fused images.(2)Application of Multi-focus Image Fusion in Fiber Identification and Fibers Contents TestingRelying on the image fusion algorithm,seveal feature paramaters of various fibers are gained by image processing methods.The paramaters are imported into the Support Vector Machine(SVM)classifier to achive the identification analysis.The recognition accuracy rate is 100% among the cotton fiber,viscose fiber and kapok fiber and the accuracy rate is 80.65% between wool fiber and cashmere fiber.Through the separation process of cross-fibers and fiber identificaiton,the length and category of each complete fiber are obtained.Based on the content formular,the contents of cotton/viscose and cotton/kapok fibers are tested.In the wool/cashmere fiber testing,except for the image processing method,the Near Infrared Spectroscopy detection is also applied to supplement the wool and cashmere fiber testing.Through the spectroscopy analysis and stoichiometry methods,the two fibers are detected.(3)Application of Multi-focus Image Fusion in Yarn Hairiness and Nonwovens Straight-through Porosity Distribution DedectionAccording to the image fusion algorithm and the image mosaic technology,a certain length of yarn fused image and certain vision field of nonwovens image are obtained.We set two sets of images-focal images and fused images for comparison,and use the same image processing technology to get yarn hairiness information and non-woven straight-through pore distribution information.In the yarn hairiness testing experiment,the hairiness information of the multi-focal yarn image is partly missing and partly integrated into the yarn trunk.In contrast,the fused image remains completely yarn hairiness and shows clear yarn edges so that yarn hairiness and yarn core are separated completely.The experimental results show that the hairiness value calculated by the fused image is larger than that of the multifocal image,and the diameter of the yarn main body by fused image is smaller.In the comparative experiment of non-woven pore distribution,the edge of the fiber in the multi-focal image is blurred and the separation between the pore and the fiber target is difficult to implement.As a result,the detection result of the multifocal image has larger pore area and smaller number of pores.Through the statistical test,the results obtained by the multi-focal images and the fused images have significant differences.The fusion images are clear and complete,and the detection results are more accurate.Therefore,in the use of image processing to detect yarn and nonwovens,in order to enhance testing accuracy,it is necessary to use the image fusion technology.In this thesis,the regional gradient variance image fusion algorithm is integrated into the application of fiber,yarn and non-woven fabric testing to realize the fiber identification and fiber contents,yarn hairiness and nonwovens pore distribution.The comparison between focal image and fused image confirms that proposed image fusion method can improve the accuracy of textile digital detection and is an indispensable part of textile digital testing.