A real-time self calibration approach for tri-axis magnetometers is proposed.The parameterized model of magnetometer readings is established by analyzing the nature of sensor errors

which lays a basis for the derivation of a nonlinear state-space model of calibration parameter using the invariability of local geomagnetic field.The real-time estimation of calibration parameter is derived based on extended Kalman filter(EKF) solution

the U-D decomposition filter is given for ease of implementation.Compared with the batch approaches

the proposed method not only can achieve real-time estimation of calibration parameters

but also is convenient for embedded implementation.Simulation results demonstrate the correctness of the derivation of the proposed approach.The effectiveness is validated by experimental test using RM3000 tri-aixs magnetometer.The results of simulations and experiments are analyzed and concluded in comparisons with TWO-STEP method.