The Technology Research Of Automatic Grasping Based On Depth Vision

As an important task of the robot,researchers have made many achievements in the automatic grasping.However,most of methods have a high requirement for the working space.In this paper,we present a method of automatic grasping,which can cope with a complex environment.This method is based on depth sensor,which can finish the grasping task by the user's instruction.The main work of this paper are as follows:(1)Use the depth images to reconstruct the working space.Using the angel data of the robot and the ICP algorithm,our method will reconstruct the working space in real-time.In addition,it can obtain the size of the aim object without making any tag on it.(2)Extract the mesh data of each aim object by using the clustering algorithm.This paper uses the method of plane fitting and hierarchical clustering to analyze the working space mesh,which can extract the mesh of each aim object effectively.(3)Obtain the grasping position by primitive shape,considering the reality that most of object can be composed of the primitive shape.This paper presents an alogorithm to compute the grasping position based on primitive shape.Moreover,this algorithm can obtain the grasping position of most of the objects.Finally,the method will be verified by the automatic grasping experiment,and the result shows that the method can effectively accomplish the automatic grasping task in the complex environment.