Abstract: The traditional local linear optimization method is widely used in the inversion problem to explore the distribution of stratigraphic heterogeneous structures and their physical properties. To address the challenges caused by nonlinearity and multiple solutions, the improved Particle Swarm Optimization (PSO) algorithm, a global nonlinear optimization technique, is employed for parameter search optimization. The strong dependence of inversion on the initial model is avoided by this algorithm. In addition, the fixed inertial weights are optimized to be adaptive inertial weights based on fitness changes. Therefore, the global optimization ability and local fine-search ability of the particle swarm are significantly enhanced. In order to optimize the forward modeling process, the Indirect Boundary Element Method (IBEM) is employed to reduce computational dimensions, and computational efficiency and accuracy are greatly improved. Considering the elliptical cavity and inclusion in the half-space, the availability and stability of the inversion based on the improved PSO algorithm and IBEM are investigated. The position of the heterostructure can be quickly inverted by the PSO algorithm improved, and the accuracy of all parameter inversion results can be highly guaranteed. Moreover, the lack of prior information is compensated by expanding the search range of the algorithm, and this algorithm can be effectively applied to the inversion of elliptical heterostructures.