Aiming at the low navigation precision of the low earth orbit satellite caused by the low precision of the geomagnetic field model, an autonomous navigation method for low earth orbit satellite is proposed by fusing geomagnetic, sunlight and spectral redshift information. The federal adaptive unscented Kalman filter (FAUKF) autonomous orbit determination scheme is designed based on geomagnetic/spectral redshift/sunlight information. The cosine of the included angle between the sunlight vector and the geomagnetic vector, the intensity of the geomagnetic field and the spectral redshift information are taken as the observational quantities to estimate the speed and position of the satellite. The simulation result is the position accuracy of 465.38 m. The accuracy of the algorithm is higher than pure geomagnetic navigation, the convergence and stability of the filtering is good, the navigation error does not accumulate with time, and has engineering application value.