Dexterous Hand Grabbing Posture Based On Stereo Vision Guidance

Multi-finger dexterous hands with multiple degrees of freedom,multi-sensing functions and good grasping operation functions have received extensive attention.Due to the diversity of dexterous hand grasping modes and the complexity of unstructured environments,the grasping planning of the target object becomes very complicated,and the grasping of the dexterous hand is largely based on the external shape of the target object,so it is very important to establish a grab system based on the analysis of target shape.Aiming at the above problems,taking the grasping posture of the dexterous hand as the research goal,focusing on the two major problems of the dexterous hand's visual perception of the environment and the determination of the grasping posture,an autonomous grasping method based on stereo vision guidance and simplification of the target object's shape.Mainly completed the following three aspects of research work:Firstly,point cloud preprocessing and target model shape analysis.The method based on statistical analysis removes outlier noise,and estimates the point cloud normal vector based on PCA algorithm.Simplify irregular complex objects into a combination of some basic shapes,use the concave-convex segmentation to divide the 3D data points of the captured object into various blocks,and fit each part into a sphere,cylinder or cuboid.Secondly,object pose estimation model based on feature description and ICP algorithm.Extract the point cloud feature points and feature descriptors of the object to be captured,and establish a feature template set,use the SAC-IA algorithm to initially register the point cloud of the object to be identified with the template point cloud,and then use the ICP algorithm based on K-d tree for fine matching.The average registration degree of the final registration reached 94.1%,and the accurate 6D posture of the object to be grasped was obtained.Finally,Determination of grasping posture of dexterous hand based on body analysis and force closure.Use the D-H parameter method to determine the kinematics model of the dexterous hand,use the hyperellipse to describe the grasping plane qualitatively,analyze the grasping feature plane of the object;perform the grasping analysis on the simplified object based on the stable grasping condition of the force closure According to the grasping force spiral space analysis,the grasping quality of the dexterous hand is defined,and the optimal grasping posture is determined according to the evaluation index.This article uses stereo vision to perceive the posture of an object.From the perspective of analysis,it establishes an accurate grasping model of dexterous hands,simplifies and fits the target model into a combination of basic shapes,and discards the complex machine learning framework,effectively reducing dexterity the complexity of the hand grab calculation.The stereoscopic vision-based grasping gesture determination method for dexterous hands proposed in this paper has a positive effect on the grasping of dexterous hands.