Micro CT Experiment And The Method Of Yarn Segmentation In Plain-weave Composites Of Micro CT Images

This dissertation,focusing on the application of composite in the structure of hypersonic vehicle,systematically explores the Micro CT experiment of plain-weave composite,microstructure analysis and yarn segmentation method.The imaging principle of Micro CT is illustrated,and the factors affecting qualities of Micro CT images are then analyzed.Two methods,yarn recognition method based on grey level and morphological characteristics of images and yarn segmentation method based on edge detection used NSCT(Non Subsampled Contourlet Transform)domain,are expounded lastly.The results of this study are the basis of quantifying and statistically analyzing microstructure characteristics of composite and constructing the relationship between microstructures of composite and macro mechanical properties.Two types of imaging principles,based on absorbing contrast and phase contrast,are introduced in the dissertation.Combining with the characteristics of plain-weave architectures,the chosen method of direction of Micro CT slice is shown.After acquiring Micro CT images of composite,the factors affecting qualities of Micro CT images are analyzed in aspect of X-ray tube voltage,artifacts and so on.Two polymer(E-Glass/Epoxy and C/Epoxy)and one ceramic matrix plain-weave composite(C/SiC)are manufactured by vacuum infusion and Polymer Infiltration and Pyrolysis(PIP)process,respectively.Micro CT images of these composites with different resolutions are then acquired by chosen appropriate parameters of Micro CT equipment.Analyses of microstructures characteristics in the preprocessed Micro CT images of three composites are then accomplished by statistical method of gray values.The results show that blur yarn boundaries within orthogonal yarns,and within adjacent parallel yarns are existed in Micro CT images of plain-weave composites.To recognize yarn intersecting boundary within orthogonal yarns,a method based on grey level and morphological characteristics of images is developed,which consists of three levels characteristic combining.Texture characteristic recognition to the result of first gray threshold segmentation is first step.Then,conducting detailed gray threshold segmentation to the result of first step.Lastly,the recognition results of orthogonal yarns are obtained by morphological correction.The examples of two typical composites show that this recognition method of orthogonal yarns can obtain accurately warp and weft yarns.To segment two yarns with internal blur boundaries caused by compressions,a yarn segmentation method based on edge detection used NSCT domain is studied in this dissertation.The boundaries of adjacent parallel yarns are extracted by NSCT edge detection method.Rectangle window is utilized to define the domain of weft yarns.Whether two weft yarns are compressed together is determined by the relationship of duty cycle in rectangle window and aspect ratio.Harris corner point detection algorithm is then used to recognize corner point on yarn boundary and determine whether each corner point is inflection point.The line through two inflection points whose distances summation to the mean axis of rectangle window is least is considered as the result of internal boundary of two adjacent compressed weft yarns.This work is important to automatically recognize and segment yarns in Micro CT images of plain-weave composites.Additionally,it is the basis of constructing the model of microstructures of plain-weave composite and macro mechanical properties of composite.The methods to analyze microstructure characteristics and recognize yarns of plain-weave composite are also suitable for general woven composites.