Fast segmentation of RGB-D images for semantic scene understanding

As robots move away from pre-programmed action sequences in controlled laboratory setups towards complex tasks in real-world scenarios, both the perception capabilities of these systems and their abilities to acquire and model semantic information must become more powerful. In this context, fast means for pre-processing acquired sensory information and segmenting task-relevant regions are an enabling technology and a prerequisite for avoiding longer delays in sense-plan-act cycles. We present (on a poster and in a demo) two fast segmentation methods for RGB-D images.