Selective laser melting is a commonly employed additive manufacturing technique that facilitates the fabrication of intricate geometries through the laser-induced melting of powder materials. The quality of the produced parts is significantly influenced by the molten pool morphology, which is affected by parameters such as laser power, scanning rate, and powder characteristics. However, the selection of unknown parameters within the heat source model significantly impacts the simulation outcomes and must be carefully considered. This study addresses this issue by proposing an inversion method for accurately determining the parameters of the Goldak double ellipsoid heat source model using molten pool morphology as a reference. A pattern search algorithm combined with Bayesian inference was employed to invert and estimate the heat source parameters. The results demonstrated that the inversed parameters significantly improved the prediction accuracy of molten pool geometry. The inverse parameters χ₀, χ₁, and χ₂ were 1.17, 1.00, and 2.08, respectively. The study provides valuable insights into the use of image-based methods for parameter inversion and offers a more reliable tool for improving the precision of simulations. These findings have important implications for optimizing processing conditions and enhancing the overall quality of additively-manufactured components.