Object Recognition Method for Home Robot Using 3D Model Data Including Grasp Area and Degree of Freedom Information of Household Items

Abstract

In order for a robot to handle household items, it is necessary to grasp the kind, position, shape, etc. We utilize the object recognition method based on detailed 3D shape models. However, the 3D model is only shape information and there is no structure information. Therefore, it is difficult to recognize the variable part depending on DOF of household items. In this paper, when using the detailed 3D model matching method using Point Pair Feature (PPF), we create model data with grasp areas and DOF information on household items required for operation and propose an object recognition method of home robot with it.