Research About Mobile Robot Binocular Vision Technology Base On DSP

Binocular vision simulate human eyes, using two cameras capture the same scene atthe same time. Due to different position of two camera, image existence disparity, so thedepth of space objects can be obtained by disparity. After years of research, thebinocular stereo vision technology has made great development, and have goodprospects in mobile robot, traffic monitoring, industrial detection, aerospace andaviation etc.Traditional binocular vision are realized in PC, in practical applications, such asaircraft, mobile robots often requires system with flexibility and portability, but thedisadvantage of PC such as large volume, high power consumption, common computecapabilities, limited research and application of binocular vision technology. With thedevelopment of embedded technology, DSP technology is more and more mature,compute capabilities improve fast, which meet binocular system high operation speed,small volume, low power consumption, stable and reliable demand. This thesis is basedon DSP to realize binocular vision system.This thesis main research content is:(1) Firstly, the system function and the overall framework is introduced, and theapplication on the mobile robot is briefly described.(2)According to the application requirements on the mobile robot, binocular visionhardware system is designed, which includes binocular camera selection and fixedstructure, the DSP board selection and peripheral support, display environment support.(3) The establishment of mathematical model of binocular vision, calibration ofcamera parameters using Matlab and OpenCV on PC, and the realization of dynamicprogramming algorithm(DP) and semi-global stereo matching algorithm (SGM),andthen obtain the3D point cloud for distance measurement and3D reconstruction bydisparity map.(4)According to the principle and development environment of DSP, configurehardware and software; including adding real-time operating system DSP/BIOS, FAT32file system, embedded vision library EMCV support, and achieve the two cameras capturing, processing and display at the same time.(5)Transplanted stereo matching algorithm DP, SGM to DSP, and improve thereal-time performance by DSP optimization.(6) Experiment and analysis of binocular vision in the detection and location of theobstacle.