A hybrid vision system for meaningful landmark isolation and recognition

Vision-based navigation is the primary motivation for the research effort narrated in this paper. In order to achieve reactive and reflexive motion for navigation, a robot must be equipped with a sensory system capable of performing rapid analysis of sensory data such that the controller is fed with current information about the environment instantly. The information acquired need to be segmented into appropriate divisions and the segmented information need to be recognized properly. The authors´ architecture for such a vision system separates these two major tasks-segmenting image information and recognizing the segmented information. There are two systems presented-one for region transmission in the global image, and the other for a neural network-based learning mechanism