High-resolution Image Reconstruction Algorithm Of Neutron CT Based On Sparse Sampling

The mobile-small-accelerator-based neutron CT with low cost,high security and and flexible deployment is becoming an important research direction.However,the neutron source intensity of the mobile small accelerator is relatively low,which has higher requirements on the reconstruction algorithm of neutron CT.In accordance with the differentiated requirements of reconstruction algorithms for neutron CT in different application scenarios,this paper carried out research on high resolution image reconstruction algorithms of neutron CT based on sparse sampling.The main research contents,results and innovations are as follows:For the detection of samples with simple structures,the adaptive regularization expectation maximization reconstruction(AREM)algorithm was presented.The effects of statistical noise on the reconstruction results were suppressed by introducing a neutron statistical noise model during the reconstruction process;a hybrid regularization method was proposed on the basis of prior knowledge of gradient image sparseness and image local monotony to constrain the reconstruction process of neutron CT,which largely eliminated noise and reduced the number of required projections;and the adaptive iterative process of the algorithm was realized by the consistency-controlled steepest descent method and the total variation(TV)change method.The test results showed that under the conditions of sparse sampling and different statistical noise levels,the AREM algorithm had better anti-noise performance and reconstructed image quality for samples with simple structures.For the detection of samples with complex structures,the adaptive regularization hybrid filter back projection operator iterative reconstruction(ARIFBP)algorithm was presented based on the AREM algorithm to improve the resolution of reconstruction images.the ARIFBP algorithm used a filter back projection operator combined with a simultaneous iteration framework to replace a statistical iteration framework in the AREM algorithm,so it colud solve the problem of low detail fidelity in the reconstruction algorithm based on TV regularization.The test results showed that under the condition of sparse sampling,the ARIFBP algorithm has better detail resolution and reconstructed image quality for samples with complex structures.Finally,the proposed algorithms were comprehensively tested by the neutron CT simulation data of Pb-CH2 models and irradiated fuel assembly models,and the neutron CT experiment data of clock models.For the Pb-CH2 models with simple structure,AREM algorithm in reconstruction quality was better than ARIFBP algorithm and other algorithms,and its reconstruction results using 25 projections was better than the reconstruction results of FBP algorithm using 803 projections.Thus AREM algorithm was more suitable for the reconstruction of samples with simple structure.Under the condition of sparse sampling,ARIFBP algorithm has better detail resolution ability compared to AREM and other algorithms;it could clearly distinguish the preset subtle structure abnormalities in the irradiated fuel assembly models.Thus ARIFBP algorithm was more suitable for the reconstruction of samples with complex structure.Based on the neutron CT experimental data of sparse sampling,ARIFBP and AREM algorithms both could achieve good reconstruction results for the clock models with complex structure;ARIFBP algorithm achieved higher resolution than AREM algorithm,thus was more suitable for the reconstruction of clock models.These tests proved that the two algorithms proposed had high practical value.