Particle Filters (PFs) have been successfully employed for monocular 3D model-based tracking of rigid objects. However, these filters depend on the computation of importance weighs that use sub-optimal approximations to the likelihood function. In this paper, we propose to enrich the filter with additional refinement steps to abridge its sub-optimality. We test the proposed approach in two different types of PFs: (i) an Unscented Particle Filter (UPF), and (ii) the recently proposed Unscented Bingham Filter (UBiF). These filters are applied to the outdoor tracking of a fixed-wing Unmanned Aerial Vehicle (UAV) autonomous landing in a Fast Patrol Boat (FPB), tested in a simulated environment with a real sky gradient filled with clouds. The use of the refinement steps significantly improves the overall accuracy of the method.