Aiming at the modeling and control problems of the continuum robot due to its innate nonlinear dynamics and infinite degrees of freedom, an improved modeling method is proposed to avoid singularities, which produce pathological behaviors mainly on straight configuration. A novel control strategy that can effectively reject external or internal disturbance is designed. First, kinematics and dynamics models of the continuum robot are built based on a proposed improved parameterization, which can solve the motion singularity. Then, a reduced order linear extended state observer is designed to observe and compensate the uncertainties and external disturbances of the inaccurately modeled system, and reduce the peak phenomenon. Finally, a tailored controller is designed to form a control loop. The convergence of the observer is proved based on the Lyapunov theory. Compared with the aforementioned method with the sliding mode control, the experimental results show that the proposed method can better integrate prior knowledge, without wasting model information, and achieve stronger robustness and disturbance rejection capabilities.