Object semantic map representation for indoor mobile robots

Abstract

We propose a novel semantic map representation method of indoor environment for mobile robots. The semantics is based on objects, which can be a desk, a wall, a room, or other things. The environment model is composed of object identifiers, object properties and relationships among them. The map is represented by Web Ontology Language (OWL) to share semantic knowledge with human. In order to extract the semantic information, plane extraction, object recognition and region inference are implemented by using the stereo image data. The experiment of representing an indoor scene shows that the method is feasible and effective to describe indoor environment.

Extracting semantic indoor maps from occupancy grids

“…A second group is formed by approaches assigning semantic labels to parts or objects of the perceived structure of the environment, like traversable terrain, trees or similar structures in outdoor environments or walls, ceilings, and doors in indoor settings [14,15,16,17,18,19,20,21,22].…”

“…Subsequently, we normalize the weights a r (s), calculate the cumulative probability and draw the second room, as done in (16) to (18). Once the two rooms are obtained, we try to combine them into one room.…”

“…Using the weight ω g , one door candidate is drawn from the set of all detected candidates, as done in (16) to (18). Then, the MH algorithm decides whether this door will be accepted.…”

Recognizing Furniture Supported by Legs Based on Vision Sensor

“…The semantic maps [1,2] come up to capture the human point-of-view of robotic environments, which enable the high-level and more intelligent robot development and the human-robot interaction as well. To realize the above function, semantic map should include detailed information about the functional properties of rooms and objects [3][4][5]. For example, if a man needs a "chair", the robot should find the semantic symbol and the corresponding location of the chair in its semantic map at first.…”

Online semantic exploration of indoor maps

“…A second group is formed by approaches assigning semantic labels to parts or objects of the perceived structure of the environment, like traversable terrain, trees or similar structures in outdoor environments or walls, ceilings, and doors in indoor settings [10], [3], [13], [17], [18], [9], [14].…”