The high-performance attitude controller can effectively suppress comprehensive effects caused by uncertainties and external dynamic disturbances, and ensure that the quadrotor can safely and reliably fly to complete the designated mission. Therefore, nonlinear uncertainties of quadrotor flight systems are approximated by a radial function neural network (RBFNN) quadrotor, and an extended state observer is designed to estimate lumped disturbances caused by RBFNN approximation errors and external disturbances in this paper. The black box problem of RBFNN is solved by using a model identification error and a tracking error as decision variables. Then, an adaptive robust anti-disturbance attitude tracking controller is designed for quadrotor flight systems based on dynamic surface control and Lyapunov stability theory, and an auxiliary system is constructed to suppress the effect of filtering error on the closed-loop system performance. Finally, Simulation results show that the quadrotor can precisely track the desired attitude angles, and the proposed controller has strong robustness and stability in the presence of uncertainties and disturbances.