Robust Algorithms for Tomographic Reconstruction of Line-of-Sight Thermal Flow Field Images

Abstract

Listen

Translate article

Publisher Summary Tomographic image reconstruction is analogous to a function optimization for the reconstructing field under a constraint specified by a type of projection. This chapter focuses on a new tomographic image reconstruction technique that uses Genetic Algorithm (GA), which is a robust, intelligent, and combinatorial function optimization algorithm that is used for finding unknowns from a set of ‘knowns' or constraints based on the mechanics of the genetic principle. The parallel optimization feature of GA allows simultaneous optimization of multiple parameters, while most conventional tomographic techniques reconstruct images for a single-parameter, such as in algebraic reconstruction technique (ART). Most optical visualization techniques measure line-of-sight projected images that integrate optical information along the beam path occupied by a test field. Unless the optical information is uniformly distributed along the beam path, a sophisticated reconstruction process is necessary to recover the cross-sectional field information from the projected images. Such procedures are generally called optical tomography or tomographic image reconstruction. This chapter discusses two conventional tomographic reconstruction schemes: Fourier Transformation Technique (FTT) and ART and multiplicative algebraic reconstruction technique (MART). The GA-based evolution optimizes the parameters of the basic functions under the constraint in that the virtual projections approach the measured projections. Despite its robustness and more accurate reconstruction potential, the GA-based tomography requires an excessive computational time because of its massive combinatorial handling of multiple solution candidates. In an effort to accelerate the calculation procedure without sacrificing the advantages of the GA-based tomography, the chapter illustrates a hybridization technique for GA evolution with a downhill Simplex optimization method.