The visual tracking system is composed of an object moving around the work space and a robot tracking the object based on the visual sensory feedback. This paper proposes a control theoretic formulation for 3D feature-based visual tracking system. The system is modeled by a nonlinear differential equation that incorporates the dynamics of the robot as well as the characteristics of the camera. A nonlinear control scheme which strictly linearizes the system is proposed. Simulations of 6 degrees of freedom control with PUMA 560 manipulator are carried out to evaluate the effectiveness of the proposed nonlinear approach.