To improve flight performance of small rotary-wing unmanned aircraft under complex environment, a composite control method based on internal model control (IMC) and adaptive radial basis neural network (RBFNN) is proposed. With the analysis of the characteristics of system disturbance, an IMC system is constructed to eliminate system errors. Furthermore, an adaptive RBFNN without prior training is proposed to eliminate residual estimation errors to augment the control performance. The effectiveness of the composite control method is validated by a series of flight tests. Compared with the feedback control method, the composite control method can yield good tracking performance under wind disturbances.