Font Size: a A A

Key Technologies For Real-time Recognition And Grasping Of Moving Objects Based On Binocular Stereo Vision

Posted on:2021-12-24Degree:MasterType:Thesis
Country:ChinaCandidate:H XuFull Text:PDF
GTID:2518306107966879Subject:Mechanical engineering
Abstract/Summary:PDF Full Text Request
With the development and improvement of automation industry,the production and processing methods are increasingly diversified and intelligent.Binocular stereo vision processes the external information by imitating human eyes with high efficiency and precision,which is a very important branch of machine vision.It has great application value in many fields such as industrial automation production,robot navigation and aerial survey.Based on binocular stereo vision,this paper studies the key technologies of online recognition and grasping of moving objects on the conveyor belt.Aiming at the identification and pose estimation of target workpieces,this paper proposes a feature description method based on normal vectors to extract key points and characterize the point clouds.Then use the method based on Hough voting to obtain the initial pose estimation of the target workpiece,and take another step to use the ICP algorithm to accurately estimate the pose of the target workpiece.For the problem of position detection and grasping of moving workpieces,the first is to analyze the position change and coordinate transformation relationship in the task;then study the calibration method of the conveyor belt,detecting the position of the target workpiece with the motion information provided by the conveyor encoder;finally,correct the final gripping position of the moving objects and solve the execution matrix of the manipulator.Based on the above content,the grabbing experimental scheme of this subject was designed.First,build an experimental platform for online recognition and grabbing of moving objects based on binocular stereo vision,and introduce the software and hardware system of the experimental platform separately,then verify the feasibility of the method described in this paper through a specific experimental case.Experimental results prove that the method described in this paper can accurately identify and grasp moving objects,which has certain application value in the field of practical engineering.
Keywords/Search Tags:Binocular vision, Feature extraction, Ttarget recognition, Pose estimaton, Dynamic grasping
PDF Full Text Request
Related items