In the magnetic interference environment, the yaw angle of the aircraft calculated by the magnetic sensor will deviate. Aiming at the estimation error caused by the magnetic interference, a magnetic sensor yaw angle diagnostic filtering method based on nonlinear extended state observer is proposed for the aircraft system with model uncertainty. Firstly, a two-parameter nonlinear extended state observer is designed to reduce the complexity of parameters. Then, based on the observer, an improved observer residual method is designed to diagnose and filter the magnetic interference data. Finally, an improved complementary filter based on error prediction is designed to further suppress the magnetic interference through fusion filtering. The static simulation results show that the data matching rate of the diagnosis method designed in this paper is more than 94% , and the fusion filtering result is phase ahead and smoother. The dynamic flight experiment results show that the diagnostic filtering method effectively suppresses the influence of the magnetic interference on the yaw angle estimation of the four-rotor aircraft, and enhances the stability and anti-magnetic interference ability of the aircraft.