Font Size: a A A

Research On Target Location And Hand-Eye Calibration For 3D Bin-picking

Posted on:2021-03-07Degree:MasterType:Thesis
Country:ChinaCandidate:Y S LinFull Text:PDF
GTID:2428330623974851Subject:Electronic Science and Technology
Abstract/Summary:PDF Full Text Request
With the gradual improvement of industrial production automation,workpiece sorting,as an important part of industrial production,is developing towards intelligence,efficiency and accuracy.The bin-picking based on 3D vision can effectively improve the sorting efficiency,free labor,and reduce production costs.In this paper,the bin-picking is different from traditional robotic grabbing based on 2D and 3D vision.It sets up the vision system with a 3D line laser sensor,and the so-called Eye-to-Hand system.This paper focuses on researching 3D vision and hand-eye calibration in practical application environment,which are key technologies to achieving bin-picking.Meanwhile,relevant experiments are carried out.The main research contents of this paper are as follows:Firstly,the paper proposes a method of getting target point cloud,based on domestic 3D line laser sensor LV800.According to the basic principle and the drawbacks of this sensor,two methods are used to optimize the origin data,which include denoising based on mean filtering and data filling based on neighboring points.And it carries out a point cloud generation method based on height data,which converts the original data into 3D point cloud.Combined with the practical application environments,a simple and fast method is used to separate the target point cloud from the scene.Then,the outlier point removal method based on neighborhood radius and point cloud downsampling based on voxel grid,are used to preprocess the point cloud.Secondly,the paper implements a positioning method of target workpiece,which is based on template matching.Point cloud segmentation based on region growing is used to segment the target into blocks.Then,the rough matching algorithm based on RANSAC(Random Sample Consensus)is proposed,which applies PPF(Point Pair Feature)to achieve matching.After several iterations by random sampling,the rough pose of target workpiece is obtained.Taking this pose as the initial matrix,the precise pose of the target workpiece is obtained by using the ICP(Iterative Closest Point)algorithm.Through experiments,the paper analyzes the positioning effect of the target workpiece,and verifies the effectiveness of the algorithm.Finally,according to the fixed transformation between the sensor and the robot in Eyeto-Hand system,the paper proposes a hand-eye calibration method based on a threedimensional calibration object.Grabbing the calibration object,the robotic arm performs multiple movements in space.Then,the mathematical model of = is established.The traditional linear solution method and the quaternion-based solution method are used to solve the transformation relation matrix .Using the error model,the paper analyzes the factors affecting the accuracy of hand-eye calibration,and proposes various methods to improve the accuracy.Through hand-eye calibration experiments,hand-eye calibration parameters are obtained.Combined with the calibration parameters,the paper transforms the pose of target into the robot coordinate.And it lays the foundation for the following practical application of bin-picking in combination with industrial robot control.
Keywords/Search Tags:Bin-picking, 3D sensor, Pose estimation, Eye-to-Hand, Hand-eye calibration
PDF Full Text Request
Related items