Object Recognition by Supported by User Interaction for Service Robots

This paper describes an interactive vision system for a robot which finds an object specified by a user and brings it to the user. The system first registers object models automatically. When the user specifies an object, the system tries to recog-nize the object automatically. When the recognition result is shown to the user, the system interacts with the user by speech. The user may help with providing additional infor-mation such as which object is the correct one (for example, ”I want the left object”), the relative position of the object (for example, ”It exists on the left of object A”, ”It exists be-hind object B”), or what the system mistakes the object for (for example, ”You mistake object A for object B”). Then, according to the advice, the system recognizes it again. Ex-periments with real refrigerator scenes are shown