Research On Core CT Image Reconstruction Algorithm Under The Condition Of Incomplete Projection

In the field of core pore percolation experiments and numerical research,in order to explore the fluid flowing or reaction process in core pores,the CT scanner is required to perform multiple rapid imaging of the core while the fluid is flowing or reacting.For conventional CT scanner,the time to scan a core sample is too long(usually tens of minutes to several hours),that is extremely unfavorable to the research of the fast fluid flowing and reaction process in the pores.In order to solve this problem,the incomplete projection method is adopted in the scan process,that is,reduction of the amount of projections and the projection exposure time,to reduce the acquisition time and improve the time resolution.While using the method of incomplete projection to improve the time resolution,it is the most important to research the core CT reconstruction algorithm to avoid damaging the quality of the reconstructed image as much as possible.In the thesis,we research the core CT reconstruction algorithm under the conditions of reducing the amount of projections and reducing the exposure time,and discuss the performance of the reconstruction algorithm under these two conditions at the same time.The specific work content is as follows:1.For the non-reactive dynamic process that does not react with the core when the fluid flows in the core pores,a reconstruction algorithm based on the weighted back projection and Aw TV constraint is proposed to accurately reconstruct the core CT with a reduced amount of projections.The algorithm uses the high-quality pre-scan reconstruction results as prior information,embeds the back-projection weight module to highlight the pore area,and uses the Aw TV model as a constraint item to preserve the pore edge.For the convergence process,the ordered subset is introduced to speed up the convergence process.The experimental results show that this method can better reconstruct the non-reactive dynamic process under the condition of the reduction of the number of projections.2.For the process that the fluid will react with the core such as corrosion and dissolution,a reconstruction algorithm combining guide image filtering and active contour is proposed.The static pre-scan reconstruction result is used as the guide image.The active contour model is used to capture the changing pore contour edge,and then the pore change area is obtained.A guide image update strategy based on pore statistics is proposed to update the guide image,and then the reconstruction based on the guide image filtering is performed.The experimental results show that this method can better reconstruct the reactive dynamic process under the condition of the reduction of the number of projections.3.For the reconstruction process under the sampling condition with the reduction of exposure time,a projection domain denoising algorithm based on the discriminative feature representation of dictionary learning is proposed to effectively denoise the projection data.A composite dictionary with distinguishing feature atoms is introduced to perform feature decomposition based on 3D representation for the sinusoidal structure component and noise component of the projection data.In the dictionary training phase,the composite dictionary is constructed using the static pre-scan projection data before the dynamic experiment.And use the composite dictionary to denoise the projection data of the dynamic process under the reduced exposure time,and finally obtain a better denoising effect.Finally,the projection domain denoising algorithm is combined with the proposed reconstruction algorithm to analyze the reconstruction quality under the conditions of low projection quantity and low exposure time at the same time.