With the advancement of minimally invasive surgery, there has been increasing interest from both industry and academia in developing flexible serpentine manipulators for surgical robotic systems. Due to internal friction and external disturbance when interacting with complex environment, position and shape of the flexile manipulator cannot be estimated solely by kinematic models. Hence, shape tracking of such manipulator is crucial to close the control loop and to study the robot models. In this paper, we propose a marker-based method with stereo vision for tracking the shape of the flexible manipulator by extracting position information of each independent joint. The performance of the algorithm was validated by experiments and was compared with that of electromagnetics tracking method.