A novel approach for online re-calibration of Wheel Force Transducer/Sensor (WFT) is proposed in this paper. It combines the initial angle alignment and shape-from-motion calibration by taking advantage of the wheel rotation characteristics. The gravitational load of the vehicle is only served as loading constraint in the calibration. Both of the two calibration procedures can iteratively update each other to reach the desired accuracy and efficiency without requiring the redundant data collection that performs inconvenient repeated loading cycles. A two-axis wheel force sensor is modelled and calibrated with the Gravity based Method (GM), the Initial-Value based Method (IVM), as well as the Independent Shape-from-motion Method (ISM) for comparison purposes. The tested results demonstrate that the proposed GM provides an accurate, practical, and effective technique for on-line calibrating the WFTs.