Font Size: a A A

Research On Target Recognition And Guidance Technology Based On Monocular Vision

Posted on:2019-12-01Degree:MasterType:Thesis
Country:ChinaCandidate:J Z LiFull Text:PDF
GTID:2438330545987851Subject:Mechanical engineering
Abstract/Summary:PDF Full Text Request
Robot visual locating technology is the important part of automation technology in industrial,has been widely used in medical,aerospace and other fields,with the development of robot technology and the gradually mature of computer vision theory,combining visual technology and robot control technology can accomplish more and more human labor,so the robot vision technology is further research and application.On the basis of analyzing the theoretical knowledge and development status of robot vision technology;The key issues such as calibration,image processing and locating method of robot vision system are studied and applied to the locating process of the weld starting point.For the calibration problem of robot vision system,in this paper,first of all kinds of camera calibration method is analyzed,according to the needs of system calibration,using Tsai two-step calibration method to making camera calibration;in the calibration processing,the camera is fixed in the end of robots,robot control camera in different directions to gain images of fixed calibration plate to calibrate camera,combined the inside and outside parameters gained and the robot pose parameters for the calibration of the robot hand-eye relation.According to initial welding position of recognition problem,first of all,processing the welding image in the denoising method based on the anisotropy of butt image,removing noise while preserving good weldment edge information,and the methods of edge detection,area filtering to extract the edge of the weldment information,and use the hough straight line detection to extract the weldment edge line information,obtained initial welding position of the image coordinates,according to different images of the same target extraction problem,the image matching based on Forstner corner detection method to achieve the same characteristics between different image matching.In target location,is to locate a point in space by adopted the method of "one eye and two bits";control the camera by robot to take pictures of the same point in the space from different directions,The target point in space goes through the same line as the origin of the camera coordinates,transformed the linear equation in the camera coordinate system into the coordinate of the base coordinate of the robot,calculating the intersection of the two lines that we can finding the 3 d coordinates of the target point in the robot base coordinate system and then control the robot guidance,this paper made a deeply analysis and research about the method of "one eye and two bits",at last,the experimental proved that the method can realize precise localization of target point.For six degrees of freedom robot guidance problem in a specific target point on the plane,this paper first adopts the method of "one eye two position" to calibrate the pose parameters of a particular plane in the base coordinate system of the robot,Solution of simultaneous equations that A line that goes through the point of the point and the origin of the camera frame and Specific plane,so as to realize the target point coordinates of the robot base frame,thus to achieve target real-time guidance.
Keywords/Search Tags:Monocular vision, Camera calibration, Hand-eye calibration, Feature extraction, Target location, Specific plane calibration
PDF Full Text Request
Related items