Research On Material Grasping Of Mobile Manipulator Based On Binocular Vision

In modern industrial production,mobile manipulator is widely used in the grasping,sorting or assembly of goods on the production line.However,due to the shape variability and complexity of industrial products,the traditional manipulator controlled by fixed attitude is difficult to achieve flexible positioning,grasping and handling.Therefore,taking the positioning,grasping and handling of mobile phone glass substrate materials as the research scene,this paper constructs a set of mobile manipulator system based on binocular vision guidance,and proposes a three-dimensional positioning method combining binocular vision and target detection to identify,locate and extract three-dimensional coordinates of glass substrate materials,so as to finally realize the accurate positioning and grasping of materials.The main work includes:(1)According to the current positioning,grasping and handling requirements of glass substrate materials,complete the system hardware selection and design of binocular vision mobile manipulator,build the binocular vision imaging model on the mobile manipulator,complete the calibration of binocular camera,and complete the coordinate system conversion between binocular vision and mobile manipulator system according to the principle of hand eye calibration,To realize the positioning and grasping of mobile robot arm based on binocular vision;(2)For the positioning of materials,the YOLOv5 s network is used as the basic network structure to identify and locate materials.In order to improve the recognition accuracy,the attention mechanism SENet and CBAM are analyzed,the attention mechanism is embedded into the CSP structure of the backbone network,and an improved YOLOv5 s + CBAM network structure is proposed.The training effect is compared and analyzed by experiments,On the self-made material data set,the m AP can reach 99.1% and the FPS is 71.According to the shape and size of the material,the detection head is improved to detect the position of the material grab point;(3)The stereo matching algorithm is analyzed,and the effects of BM and SGBM algorithms are compared experimentally.Finally,SGBM algorithm is used to obtain depth information.Aiming at the problem of depth information acquisition error caused by stereo matching error of grab points,an improved ranging method based on gray value is proposed to obtain the distance of materials,and the average ranging accuracy is5.16 mm.Combined with target detection,Realize the acquisition of three-dimensional coordinates of material grab points.Finally,the experimental environment is built and a complete verification experiment is designed.The experimental results show that the mobile manipulator can accurately complete the identification,positioning and grasping of materials,and transport them to the target location,which verifies the effectiveness of this algorithm.