Abstract: Inspired by the architecture in some typical examples of advanced intelligent robots,a new point of view that machine vision should also be structured in levels and modules has been put forward in this paper,in view of the endless debates between defferent theories for a long time. It can be well satisfied with the requirements for the intelligent robot control to have a compromise between the general and special tasks in a vision sub-system. After discussing the detailed level architecture and functional module,an integrated vision sensor is presented,in which a synchronized spot scanning triangulation method acts as 3D imaging sensing within the near range for precision data,a laser radar plays a role at the far field to get the coarse images,and a CCD camera plays a role to get the intensity images. They are united in space and time,and all the information can be matched and complemented each other to ensure the reliability of the database and the feasibility of the architecture.