New objects seizure method in mobile robotic using a visual servoing and neural network classification

This paper describes the development of a new objects seizure method in the robotic service framework. Its main objective is to make a manipulator arm equipped with a grip and two sensors (camera and sonar) able to handle objects in a human environment. Ultrasonic information is used by a neural classifier in order to give us object recognition and distance information. In the same time, a camera takes an image of the object. We extract its edges by image processing. So it is possible to match image center and object center. The object seizure strategy uses image and ultrasonic information both. This strategy is applied to two kinds of objects: sphere and cylinder.