Localization algorithm for 3D four-wheel alignment based on monocular vision

Abstract A novel localization algorithm for a four-wheel alignment is present. As our algorithm is based on monocular vision, the problem that a traditional 3D four-wheel alignment needs to unify the global coordinate system and easy to cause error accumulation is not existing. In accordance with the principle of monocular reconstruction in computer vision, the representation of equivalent reference plane for each wheel in its corresponding camera coordinate system is calculated out. Based on these new definitions, related angle parameters of the vehicle wheel are finally determined. Simulations and analysis show the feasibility and robustness of our algorithm described in this paper. Experiment results show that the proposed algorithm can obtain precise results and the root mean square error is about 0.005 degrees. Compared with the traditional localization algorithm of a 3D four-wheel alignment, the measurement precision of our proposed algorithm is similar, but does not need global coordinate system calibration, so the operation is simple and efficient. Furthermore, as the wheel measurement is relatively independent, it is not easy to cause error accumulation.