Research On Feature Extraction Algorithms On Micro-focus CT Image Of Polymer Bonded Explosive

Polymer bonded explosive is one of the most widely used explosive materials.In the process of preparation of explosives,defects such as pores and bubbles may be formed in the process of synthesis,crystallization,granulation...;in the process of press molding,defects such as local delamination and looseness between crystals,within crystals or between crystals and binder may be formed;Due to the existence of residual stress after forming,new damage may be formed due to external load during subsequent processing,storage and transportation.On the one hand,these defects will lead to the deterioration of the mechanical properties of PBX,and eventually lead to material damage;on the other hand,as an energy gathering area may become a "hot spot" directly affecting the mechanical properties,safety performance and detonation performance of PBX.Microfocus X-ray tomography(?CT)is a widely used imaging technology in the field of non-destructive testing.It has the advantages of intuitiveness,reliability,high sensitivity and resolution,and non-overlapping images.Through the ?CT scan,the researchers can visually observe the internal structure of the PBX.However,affected by the CT imaging conditions,the physical and chemical properties of the detected object and the image reconstruction algorithm,the acquired image is easil y blurred at the subtle defects,and the resulting image will have more noise,affecting the experimental personnel to PBX physical chemistry.Nature of research.Because CT images have characteristics such as high noise,low contrast,dark field of view,and inconsistency of local brightness and darkness,the existing detection methods are used,and the manual detection speed is slow,which is greatly affected by the staff and the environment;classical image algorithm processing,such as threshold segmentation and edge Detection,phase consistency,etc.,it is difficult to extract fine defects,and the PBX structure of the reaction is not accurate enough to meet the needs of research and production.In this paper,the ?CT image with high noise,low contrast,dark field of view,local brightness and inconsistency is summarized.The digital features of the image are summarized.Two feature extraction algorithms are proposed.In the aspect of crack feature detection: combined with non-local mean filtering,The Laplacian sharpening and gamma correction method preprocesses the image,and then directly extracts the feature points of the crack by statistical methods.The local directionality of the crack feature points is determined by the Mahalanobis distance to d erive the crack propagation direction.Finally,the fine identification of the complete crack in the CT image of the polymer bonded explosive is realized by the pixel gray value copy.The contrast between the proposed method and Canny algorithm and phase consistency method was carried out on the CT images of three PBX thermal coupled loading damages.The results show that the proposed method can extract various shapes of cracks clearly and accurately,which proves the effectiveness and high efficiency of the new method,and significantly improves the recognition and characterization ability of crack features.In the detection of particle boundary features: this paper uses statistical information The image is adjusted,and then a new algorithm is proposed.The capsule algorithm improves the selection of the watershed point of the watershed algorithm,calls the watershed algorithm based on the annotation to segment,and finally uses the statistical method to postprocess the watershed result image.Experiments show that the method can clearly and accurately outline the boundaries of the crystal.