3D Object Recognition And Localization And Its Application In Robot Bin-Picking

With the development of computer technology and machine vision technology,the robot based on visual system is becoming the hot spot of people's research.Compared with the traditional robot,the robot with visual system has the ability to identify specific objects.At present,many industrial robots equipped with vision sensors can identify and locate the specified object.However,the high precision visual sensor price is generally more expensive,virtually increase the production cost,is not conducive to the popularization of enterprise applications.In addition,the traditional research on object recognition and localization mainly focuses on the two-dimensional field.Due to the lack of information in the height(depth)of the object,it can not accurately describe the shape and posture of the object.Therefore,it is of great practical value to develop a 3D object recognition and location system based on cheap Kinect vision sensor.In this paper,object recognition software and three-dimensional object recognition and positioning system are designed based on ABB IRB1200 robot and Kinect vision sensor combined with Point Cloud Library(PCL)and Qt-designer.Firstly,this paper introduces the overall scheme of 3D object recognition and localization system,and then introduces the hardware and software tools used in the research program.Then,the common key technologies of 3D point cloud processing are described,and the imaging principle of Kinect and the method of acquiring real-time point cloud combined with PCL are introduced.Secondly,the differences and advantages of object recognition based on global feature descriptor and object recognition based on local feature descriptor are introduced.Then,the construction process of object recognition model library based on VFH global feature descriptor is introduced,and the principle and process of object recognition and matching are described.After solving the problem of recognition,the method of object location is introduced.Object orientation contains two meanings:object position determination and object pose determination.As to the position of the object,the center of gravity can be calculated to replace the coordinate of the object.However,the precision of this method is insufficient.In this paper,a custom grabbing point is used to solve the problem of the position of the object The pose information of the object is the decisive factor for the success of the object grasping.In this paper,the pose information of the object is determined by the combination of object pose matrix model library construction and ICP calibration.Finally,the hardware connection between ABB robot IRB1200 and IRC5C control cabinet,stepper motor,Arduino and stepper motor driver is introduced.The robot's grasping test and the visualization of the experimental results are also implemented by using the design identification software.Finally,the experimental results are analyzed and the results of this paper are summarized.