The Research On Recognition And Location Of Work-piece And Grabbing System Based On Binocular Stereo Vision

Binocular Stereo Vision is a hot issue in the field of ComputerScience, Binocular Stereo Vision can simulate binocular vision of humanvery well, obtaining the depth information of the object in the field of viewaccording to Triangular geometry, rebuilding the3D information of objects,it is widely used in robot recognition and location, medical image analysis,agriculture, construction industry and so on, thus, researching on BinocularStereo Vision has great significance both in theory and practice. This t hesis isbased on the basic theory of Binocular Stereo Vision, using cameracalibration and image processing techniques, developing the recognition andlocation of work-piece and grabbing system, which is based on BinocularStereo Vision, and designing the three-axis grabbing platform which has TIproduced TMS320F2812as the main control chip to do experiments onrecognition, location and grabbing, the experiment results turn out, thissystem has certain value in application. Main researching content asfollowed:(1)Camera calibration, building the imaging model of camera,considering the impact of camera distortion in the process of imaging,adopting the classical Zhang Zhen-you two steps calibration, obtaining theinternal and external parameters of binocular camera, and acquiring therotation matrix and translation vector between camera and real worldcoordinate.(2)Feature extraction and work-piece recognition algorithm, because ofcertain conditions (rotation, zoom, shelter, etc.) existing in the assembl y line,adopting SURF algorithm with integral image to recognize the work-piecewhich is prepared to be grabbed, through experiments to verify the extractionresult used SURF algorithm under noise interference, using templatematching algorithm which is based on feature point extraction to recognize work-piece under different circumstances as above, through experiments toverify the effectiveness of algorithm.(3)Stereo matching and3D reconstruction of work-piece, due to thenumerous error matching points, they will influence the result of matching,and because of the image searching in the process of matching is under2Dconditions, the execution efficiency of algorithm is low, this thesis improvesthe SURF algorithm in the process of stereo matching, adopting stereomatching algorithm which is based on relative Euclidean distance andepipolar constraint to remove the error matching points, the rapidity andeffectiveness of searching, through experiments to check the analysis.(4)Designing TMS320F2812as controller, including stepping motor,encoder, ball screw, osculum type electromagnet as carriers, designingthree-axis motion platform for work-piece grabbing, transmitting the locationinformation to the grabbing system after the3D reconstruction of work-pieceto do the experiments of grabbing, analyzing the grabbing error, it is verifiedthat this system has certain value in the industrial manufacture.This thesis researches on recognition and location of work-piece andgrabbing system based on binocular stereo vision, through experimentalverification, this system performs well in real-time and operates stable,expanding the application of binocular stereo vision.