Vision System for Robotic Handling of Randomly Placed Objects

This paper presents the development of a method for the calculation of coordinates of a part's features in a world reference frame (WRF), using images acquired from a camera. The method is used to identify multiple, randomly placed, similar objects from the consumer goods industry. The geometry and the intensity differences among multiple points are used for the identification, due to the objects’ complicated geometry and the non-discrete characteristics. All the general (e.g. orientation) and special (e.g. visible side) purpose features are identified. It was implemented within a software tool using MATLAB and tested in the case of shaver handles’ feeding.