By eliminating the data offset caused by a disturbing magnetic field during the flying condition of a spinning projectile, an adaptive unscented Kalman filter is applied to estimate the actual data of a geomagnetic field and the false value introduced by a disturbing magnetic field. For the ratio methods related to geomagnetic attitude measurement, the physical significance of the ratios obtained by all ratio methods is clarified. A new definition for the ratio is presented to ensure online estimation and numerical stability. The variation in the sensor outputs along the lateral directions is suggested to be a harmonic motion with the assumption that the spindle is fixed or in slow motion. The true data employed as a reference for comparison is generated by a dynamics model for spinning objects and by projecting the geomagnetic field to the measurement frame. Signal gathering and data preprocessing are considered. A signal-disturbing ratio is introduced to indicate the proportions of the real geomagnetic signal and the interference signal in the data acquisition process. The algorithm of the adaptive unscented Kalman filter is discussed. Simulations are performed using a 1 kHz sample rate and a duration of 60 seconds. The filtering results supported by a conventional unscented Kalman filter and adaptive unscented Kalman filters are compared with actual data for the conditions of different signal-disturbing ratios and 10 shifts of the disturbing magnetic field. The results of the simulation indicate that the proposed adaptive unscented Kalman filter with the introduced state equation achieves better performance than other filters mentioned in the paper. The filter has the ability to provide clean data from the noisy data flow measured from a spinning projectile for the ratio methods based on the geomagnetic measurements.