1 May 2026 in SSRN Electronic Journal
To achieve the three-dimensional reconstruction of smoke, this paper combines data calculation and processing by taking photos from multiple angles and utilizing computed tomography imaging technology. The SART algorithm and TV regularization are selected as reconstruction algorithms. Since the semi-convergence phenomenon exists in the separate SART algorithm and the lowest point of error in the calculation process is predictable, which can be used as the entry point for the coupling of the SART algorithm and TV regularization, this paper proposes a method of dynamically adjusting the step size of the gradient descent of the TV regularization to ensure tha a lower reconstruction error is maintained during the whole process of the calculation, under the condition that the relaxation factor of the SART algorithm is kept unchanged. Comparative evaluation against established methods demonstrates that the proposed approach achieves a statistically significant improvement in accuracy and noise reduction, coupled with enhanced stability and robustness. Especially, the effect is even more pronounced when there is input noise.