In this paper we propose a tracking adaptive impedance controller for robots with visual feedback. It is based on a generalized impedance concept where the sensed distance is introduced as a fictitious force to the control in order to avoid obstacles in restricted motion tasks. The controller is designed to compensate for full non-linear robot dynamics. Robot parameters adjustment is introduced to reduce the sensibility of the controller design to dynamic uncertainties of the robot and the manipulated load. It is proved that the vision control errors are ultimately bounded in the image coordinate system. Simulations are carried out to evaluate the controller performance.