Research On Shape Recognition Of Sheet Metal Based On Flame Cutting And Development Of Its Manipulator Grasping System

At present,the machine vision has become a mainstream research direction in the robot field.Through the visual system,the processing parts can be automatically positioned and grasped in the production and manufacturing field by manipulator,the fruit can be automatically picked in the field of agriculture,and the goods can be automatically sorted in the field of logistics.Thus,the manipulator with visual system can liberate the labor force and complete the tasks that the traditional manipulator cannot complete.In many industrial scenes,the template-based matching algorithm is generally used to achieve target detection object recognition.This algorithm is simple and easy to implement,and has a high accuracy for recognizing a small number of objects in a simple background.However,for the sheet metal with many kinds,different shapes and external features,it is uncertain to grasp with this algorithm.Therefore,in this thesis,the target detection method of depth learning is adopted to realize the positioning and grasping of the plates by the manipulator.Kinematics modeling of the manipulator includes: the DH motion model of the manipulator is established and the inverse kinematics transformation matrix is calculated using the geometric solution method.The calibration of camera internal parameters is completed.The hand eye calibration system with hand out of the eye is established and the transformation matrix between the camera coordinate system and the robot hand coordinate system is obtained using the nine point calibration method.The plate center of gravity identification and grasping is realized.Six different shapes of sheet metal are used to construct deep learning sample data sets with data-enhanced operations.Combined with the requirements of grasping tasks and the characteristics of various deep learning algorithms,the mask-rcnn algorithm is selected to detect the contour of objects,to complete the positioning of the barycenter coordinates of sheets,and to realize their grasping by manipulator.Finally,experiments are carried out to verify the accuracy of the above algorithms.