||Object detection and segmentation is one of the basic competences that a domestic robot needs. Many current approaches use information about the objects for top-down detection and segmentation. However, in many applications, the robot will regularly encounter unknown objects. In that case, top-down knowledge cannot be used. We focus on the problem of bottom-up detection and segmentation of unknown objects. Since the Gestalt psychology studies the same phenomenon in human vision, we propose the utilization of a number of Gestalt principles.
In this talk, I will motivate the use of Gestalt principles. An attention model based on symmetry will be presented and I will show that it can be used both to predict human eye fixations and to detect salient objects in a scene. I will furthermore introduce models for figure-ground segmentation and for the evaluation of segments using Gestalt principles and show their application in robotic vision.