Research On Fixed Visual Control System Of Assembly Robot Based On Monocular Images

Assembly robot with visual function has more and more widely application inindustrial fields. Compared to other types of robots such as program controlled,teaching-playback robot and so forth, visual assembly robot will undoubtedly havestronger ability to adopt the environment and intelligent level. So they are moresuitable for development needs in the intelligent assembly task. Visual assembly robothas become a research hotspot in the field of robot at present.In the aspect of image detection, the paper adopts end of executor detectionmethod based on Hough transform taking into account the edition that the specialtapered structure of end executor and the corner may blur because it is out of thedepth of field of camera. The method extracts two clearer edges to figure out terminaland thus reduces the error effectively. The detection of centroid coordinate ofirregular projection of part utilizes the principle of image moments to proceed.In the matter of camera calibration, the intrinsic parameters based on translationof camera three times is researched. In the first place, the scene points are measuredby translating camera to calibrate weakly the basic matrix of camera, next the imagecoordinates of corresponding points are recorded in two images, then it combines thetranslation vector to work out the basic matrix. So the linear intrinsic parameters canbe solved by the basic matrix obtained and its properties. The nonlinear distortionparameters is calibrated by the RAC constraint and movement of camera. We need tomove camera in vertical direction and record distortion projection coordinates of aspatial point in every positions of camera, then the coordinates is put into the RACconstraint equation set to solve parameters. The method cuts down the measuredpoints and reduces calculated amount significantly.A three-stage visual servoing method based on cross ratio invariant anddisplacement of movements is proposed in the terms of visual control. Firstly it usesreference diagram to align. Secondly it utilizes cross ratio invariant and distance ofmovement to calculate the object point, then uses PID to adjust and approach. Themethod rebuilds new constraint in every images, so the focal of camera can bechanged.The paper designed a fixed visual control system based on monocular imagesmainly aiming at the task of target assembling of small parts of visual assembly robot.Firstly the thesis studies in depth on the contents of objective detection, cameracalibration and visual servoing etc, then gets the parameters of visual controller, and next establishes the prototype system aiming at small part of submilimeter level fromboth hardware and software, finally shows that accuracy and feasibility of it throughthe related experiment of guiding to a small part target and contrast experiment ofvisual servoing based on position. Compared to other system, especially themonocular vision system, it has advantages in precision.