A 3D reconstruction method of bubble flow field based on multi-view images by bi-direction filtering maximum likelihood expectation maximization algorithm

Aiming at the problem of the three-dimensional (3D) reconstruction of bubble flow field, a Bi-Direction Filtering Maximum Likelihood Expectation Maximization (BDF-MLEM) algorithm based on multi-view bubble images was proposed in this work. The calibration method for the multi-view was studied to unify the world coordinate systems and obtain the mapping relationship between the 3D world and the projection images. Combined with the calibration results, a 3D reconstruction algorithm of BDF-MLEM for bubble flow field was developed from MLEM. In BDF-MLEM, the setting method of iterative initial value was optimized and the projection and back projection methods in the iterative process was improved based on bi-direction filtering. The experimental re-sults have shown that, compared with SART algorithm and MLEM algorithm, both the reconstruction speed and reconstruction accuracy of the proposed BDF-MLEM algorithm were greatly improved. In addition, the recon-struction results of overlapped bubbles have been improved to some extent by using the proposed method.