Improved Multi-phase Segmentation Algorithm For Industrial CT Combined With Edge Detection

With the widespread use of composite materials in modern industry,practical and theoretical research into transition regions in multi-layer structures is intensifying,and the rigorous detection of their information is of great importance in advancing the development of composite materials.In order to detect transition region information in composites,the composite material can be scanned using an industrial scanner and the resulting images processed.To ensure the ultimate accuracy of detection,the accuracy of the CT image segmentation needs to be improved.Since composite materials are complex structures with multiple materials,their CT images contain both partial volume effects,noise and low contrast.Partial volume effects can lead to inaccurate measurement of CT values at the edges of multi-phase junctions in hybrid CT slices.The spatial longitudinal variation of the material can cause the grayscale values of the first end of the material in the mono-layer slice to be different from the true CT values and cannot be extracted correctly.At the same time,there are low-contrast target regions in the CT image with a small amount of noise,which increases the difficulty of accurate segmentation.Some of the current segmentation algorithms are not able to adequately solve the existing problems for composite materials.Therefore,this paper proposes an industrial CT multi-phase segmentation algorithm combined with improved edge detection.The main research work of this paper is as follows:1)The natural image segmentation algorithm and industrial CT image segmentation algorithm are summarized and analyzed,and the theory of industrial CT image segmentation methods at home and abroad is outlined,and their advantages and disadvantages are summarized.2)The X-ray tomography technology and the theory of classical image segmentation algorithms are introduced,and the basis of the algorithms used in this study is clarified based on the image characteristics of multi-phase industrial CT and the performance of various image segmentation algorithms on multi-phase CT.For partial volume effect,the paper adopts the Canny operator with superior performance in edge detection,which can accurately extract edges based on non-maximal value suppression;for unrealistic CT values at the first end of materials in single-layer slices,the paper adopts the 3D seed region growth algorithm,which can accurately segment this part of the target by combining spatial and gray-scale information.3)For the complex target region in composite CT images,there is often a problem of edge loss of Canny operator caused by low contrast,and an improved algorithm for edge detection of industrial CT images based on Canny operator is proposed.The information of low contrast is enhanced by using multi-scale detail enhancement algorithm,and then weak edge pixels and noise points are classified according to the effective features within the neighborhood of edge pixels to accurately detect the edges in multi-phase CT images,thus providing more accurate growth stopping conditions for subsequent region growth.4)Due to the characteristics of poor continuity of homogeneous targets,low contrast in some regions of liquid phase and partial volume effect in multi-phase hybrid CT images,the problems of missing target regions,over-segmentation and wrong selection of initial seed points occur when seed region growth method is used.To address the above problems,a multi-region target extraction algorithm combined with improved edge detection is proposed for multi-phase segmentation of industrial CT.The edges are introduced in the global seed region growth process and combined with the stop-growth threshold to improve the growth stopping condition of single seeds and suppress the overgrowth of local low-contrast targets.Finally,it is experimentally verified that the method in this paper can obtain better segmentation results in complex composite multi-phase CT images compared with other algorithms.