Particle Filter-based Spherical Rotor Attitude Visual Estimation

Rotor attitude estimation (RAE) is the prerequisite for the close-loop control of permanent magnet spherical motors (PMSpM). This paper presents an RAE method of PMSpMs using monocular vision. The FDSST algorithm is improved to track three visual objects which are set on top of the PMSpM rotor output shaft simultaneously. To raise the performance of dynamic tracking and anti-occlusion resistance, the particle filter algorithm is introduced to optimize the multi-object FDSST tracker. To verify the proposed particle filter-based multi-object FDSST, the one-object offline simulation is conducted on the customized dataset. The simulation result proves the feasibility of the particle filter-based visual RAE method for the PMSpM.