Research On Object Recognition And Localization And Grabbing System Based On Laser Point Cloud

This paper studies robotic grabbing technology based on visual guidance.At present,industrial robots have a large number of application scenarios,and the combination of artificial intelligence technology and traditional robot technology has become a mainstream trend to further increase automation,intelligence,and increase production efficiency.In this context,this paper deeply studies stereoscopic vision,laser scanning reconstruction,workpiece recognition and positioning,robot grabbing and other technologies,and completes the integration of these technologies,and developed a robotic grabbing system based on visual guidance.Firstly,the scene reconstruction method based on stereo vision and laser scanning technology is studied.The camera model and the stereo vision system model are briefly introduced.For the camera calibration problem,a more accurate iterative calibration method is adopted,by which the calibration accuracy is significantly improved compared with the traditional method.A laser scanning device was designed and the scene reconstruction accuracy experiment was carried out.The results show that the reconstruction accuracy of the system can reach less than 1mm,which meets the requirements of robot crawling.The method of workpiece recognition and location based on point cloud is studied.Firstly,the traditional point-pair-feature-based(PPF)method is introduced,including feature description,pose estimation principle,voting method,and clustering strategy.Then based on the framework,the improvement approaches adopted by our system are given,including the introduction of viewpoint sampling strategy,the improvement of offline training process,the improvement of voting score based on the idea of visibility,and the improvement of pose averaging method.The effectiveness of the improvement was verified using experimental methods.A robot vision hardware system was built and a corresponding visual guidance software system was developed.In this process,the hand-eye calibration of the robot vision system was completed,and the offline simulation scene of the established hardware system was built using the VREP robot simulation platform.The workpiece recognition and positioning,path planning,interference check,path production,etc.can be performed in the simulation scene.In order to verify the effectiveness of the system,two parts were selected,corresponding to two kinds of pick-up devices,and the offline picking and online picking experiments were carried out.The results show that the vision guiding system developed by this subject can stably and effectively pick and place the workpiece.