Both autonomous and teleoperated tasks with a dexterous manipulator often use cameras for external sensing. For teleoperated tasks, an array of cameras is typically used to provide the operator with multiple two-dimensional views of the manipulator workspace. For autonomous operations, cameras are used for visual servoing or to produce a map of the environment in the manipulator workspace. Nominal operations will likely produce manipulator configurations that occlude the line of sight from the camera to a target of interest. One possible approach is to treat the camera line of sight as a virtual obstacle to prevent camera occlusion. This approach is demonstrated on the Ranger dexterous manipulator for a variety of task configurations. Extension of this approach to non-redundant manipulators is also considered.