Muon tomography is a technique that aims to reconstruct the internal composition of an unknown volume by analyzing the trajectories of muons as they traverse the volume. The reference method proposed by Schultz et al. (2007) models the scattering angles and displacements of the observed muons using a Gaussian distribution and employs the Expectation-Maximization (EM) algorithm for maximum likelihood estimation. We propose an extension of this model which uses a mixture of two Gaussian distributions to better account for the heavy tails observed in the distribution of the scattering of muon tomography data. The proposed model is fitted using the stochastic EM (SEM) algorithm. An application to simulated data from an experiment aimed at evaluating the wear level in the inner walls of an insulating tube, is given.