Development Of Binocular Vision System And Research On Its Control Method

The development of machine vision technology enables machine to observe andperceive the outer world like human beings. In the1980s, Marr’s visual theorypromoted the development of the technology of machine vision further. Stereobinocular vision technology, a branch of machine vision technology, getting image bytwo camera, determining the depth of the scene with the help of parallax principle, cangenerate three-dimensional information and location information of the object afterprocessed by relevant algorithm. With the rapid progress of image processingtechniques and computer technology, the application of Binocular stereo visiontechnology become more and more widely in production and life. Based on the basictheory of binocular stereo vision, we designed a mechanical platform for the study ofthe binocular stereo vision technology and built the control system of the platform, andmore over, we did some research on the control algorithm, and finally the design ofhumanoid binocular stereo vision system was done.Firstly, the overall design scheme was proposed according to the actualrequirements of the subject after learning and researching the binocular stereo visiondevice from both domestic and overseas, and then we came up with the designspecifications, on this basis, we developed a2-DOF binocular stereo vision platform.This device with a simple structure has a small size and light weight. There are nocoupling between the two DOFs, and each DOF is driven by a brush DC servo motor.Secondly, we built the control system of the binocular stereo vision platform afterfinishing the design of the mechanical part. Considering we need control the position ofthe two DOFs, we choose a position closed loop digital servo control style for thesystem after comparing related control technology. As the system should be carried onthe bionic hexapod robot, we designed a motion control card which takes DSP F28335as its core processor. Based on above mentioned hardware system, we built themechanical and electrical system model of the two joints and simplified it, then wedecided to use positional digital PID algorithm to control the spin joint and usefeedback and feedforward algorithm to control the pitching joint, and finally simulationof the two joint’s control algorithm was done using MATLAB/SIMULIK.Finally, three experiment including starting to find zero position experiment, jointcontrol algorithm experiment and target tracking experiment was done to validate therationality of our system. And the result shows that the design of mechanical part of thesystem is reasonable, and can reach the goal of tracing moving target.