Rotor Attitude Estimation for Spherical Motors Using Multiobject Kalman KCF Algorithm in Monocular Vision

Abstract

This article proposes a rotor attitude estimation (RAE) method for spherical motors using the multiobject Kalman kernel correlation filter (MKKCF) algorithm in monocular vision. A permanent magnet spherical motor (PMSpM) is adopted as the research object, and a monocular area scan camera is equipped in the RAE system with a visual feature component mounted on top of the rotor output shaft. In the proposed MKKCF algorithm, the Kalman filter is used to enhance the robustness and accuracy of KCF tracker for each object. To simplify the proposed algorithm verification, a one-object tracking comparison is conducted among the KCF algorithm, the fast discriminative scale space tracking (FDSST) algorithm, and the Kalman KCF algorithm, and the results show that Kalman KCF tracker is more applicable for RAE. In addition, an RAE test bench is developed, and the MKKCF-based RAE method and the micro-electro-mechanical system (MEMS) (MPU9250) method are compared when estimating the rotor attitude. To set a benchmark, the contact rotor position measurement with encoders is used. The comparison results indicate that the MKKCF-based RAE method works with higher accuracy than the MEMS method. Finally, a closed-loop PMSpM control experiment is conducted by using the proposed MKKCF-based RAE method, and the practicability of the proposed method is proved.