Diffuse reflection model and noise stabilization for tangential image tomographic reconstruction (TITR) code

摘要

A tomographic reconstruction code had been reported by us for inferring the poloidal emissivity of tokamak plasma from tangentially acquired images. Here we present modifications to the code that account for any diffuse reflections from the surfaces of walls enclosing the plasma. It is generally recognized that such reconstruction codes are highly susceptible to noise in the data. In this work we have analysed the sensitivity to noise for varying degrees of over-determinism in the set of equations; over-determinism is defined as the ratio of the number of detector signals available to the grid resolution of reconstruction. A tractable scheme for dividing the poloidal cross section into a finite number of unknown sub-tori and voids, while still keeping the over-determinism high, is incorporated. Finally it is shown that noise level >20% can be handled with over-determinism achievable from present day detector array/cameras. The singular value decomposition of the matrix, as used here, can be expected to converge even if any ill-conditioned matrix is encountered due to computational round-off errors in the estimation of chord lengths through sub-tori and voids.