Research On Damage And Feature Extraction Methods On 3D CT Image Of Polymer Bonded Explosive

Polymer bonded explosive(PBX)is a kind of insensitive heat-resistant explosive with excellent performance,which has been widely used in the field of weapons.During the preparation,transportation and storage of PBX explosives,cracks,holes and other damage will occur.These damages will not only affect the mechanical properties of explosives,but also affect the safety of explosives.If a model of PBX explosive damage distribution can be established,it can help researchers understand the formation,development and evolution of damage and how to suppress the occurrence and development of damage.Moreover,TATB-based PBX explosive shaped pellets contain a certain amount of binder in addition to explosive solid particles.Relevant researchers need to calculate and simulate the performance of PBX explosives,improve the preparation method of PBX explosives,all need to be based on PBX explosive mesostructure information.This requires us to dig out the distribution of the solid particles of the explosive and the boundary of the binder from inside the TATB-based PBX explosive.Computed Tomography technology gradually plays an important role in the internal structure analysis of PBX,because of its advantages in analyzing the internal structure of materials.However,due to the limitations of CT equipment or the physical and chemical characteristics of samples,the images obtained by CT scanning have the problems of dark image field,serious noise,and low contrast.Researchers cannot directly establish a three-dimensional structure model of PBX explosives based on the original images.Unfortunately,the existing classic image processing algorithms cannot effectively extract damage or feature distribution information from images with the above problems.In response to this situation,we have proposed a series of specialized algorithms to extract CT image information of PBX explosives to meet the needs of researchers.For the damage detection part,we propose a processing algorithm for weak damage from 2D to 3D,based on Chebyshev inequality to initially extract the suspected damage,and then propose an elastic capsule scanning algorithm to filter the suspected damage to obtain a better 2D recognition results;we also propose to use the correlation between the three-dimensional slices to further reduce noise,and finally establish a visual model of the spatial distribution of damage through interactive means and spatial connected domain analysis,grayscale morphology transformation.For the PBX particle segmentation part,we used(a)elastic capsule particle segmentation algorithm,(b)watershed algorithm based on elastic capsule results,and three-dimensional correction algorithm based on the(a)(b)algorithm results;we use the three-dimensional correction algorithm to strengthen the boundary features,establish a weak boundary constraint in the original image,and combine with the Cellpose model to further improve the particle segmentation result.After the results of the above methods,the particle segmentation is more accurate,but there are still some errors that are unavoidable.Therefore,a manual labeling tool is provided to correct the wrong segmentation with a small amount of manual intervention and establish a more accurate three-dimensional particle visualization model.