Research On Segmentation Processing And Feature Recognition Method For Fiber Composite Material CT Image

In recent years,with the development of industrial fields such as bridges,ships,and aerospace,composite materials have been applied to more and more fields with their superior properties such as low density,high strength,high modulus,high temperature resistance,and corrosion resistance.The mechanical properties of fiber composite materials are involved with the shape and distribution of fibers inside the matrix.In the research on the properties of fiber composite materials,microtomography(Micro-CT)has always been an important method to detect the internal structure of composite materials.How to make full use of CT images to observe the internal structure of materials to obtain accurate fiber shape and distribution information is of great significance to the reconstruction of materials and is of great value to the study of composite material properties.Due to the limitation of equipment accuracy for fiber composite material CT images,the current feature extraction methods are often unable to obtain accurate and usable feature information.Based on the watershed algorithm,the Unet neural network model and the non-rigid registration method,an improved image segmentation processing algorithm is proposed in this research,which can be used to effectively improve the recognition of fiber features within the material,and then accurate fiber distribution information is obtained through feature recognition technology.A complete framework of fiber composite material CT image processing and analysis is finally constructed to provide guiding solutions for composite material CT image processing and analysis problems.The main research contents of this paper are as follows:(1)Aiming at the problem of unclear fiber boundaries in CT images,the morphological assessment is performed in the front sections of the fibers and the watershed algorithm is used to process them.The Unet neural network model is used to predict the image segmentation labels in the oblique section area to ensure that the image is reasonably segmented.Furthermore,an image segmentation method based on improved watershed algorithm combined with Unet model is proposed.On this basis,a precise segmentation process of composite material CT images for multiple crosssectional shapes is constructed,which provides technical support for subsequent fiber feature information extraction.(2)In view of the three-dimensional characteristics of composite material CT images,a reasonable image registration technology is selected to construct a conversion model between similar images.Based on the nonrigid registration algorithm of sequence images,an improved combined method of image segmentation and registration is proposed in this paper based on the selection of reasonable transformation models,similarity criteria,search spaces and search algorithms,which improves the precision and efficiency of image 3D features processing and provide effective feature information for feature extraction.(3)According to the acquired three-dimensional information of composite materials,the spatial characteristics of fibers are extracted through connected region analysis and RANSAC algorithm,and the spatial distribution and structure information of single fibers are obtained.In this paper,a series of processes of image segmentation processing,image registration and 3D feature detection are established,and the validity and rationality of the method are verified through application research.