Research And Design Of Automatic Obstacle Avoidance Of Embedded System Based On Monocular Vision

With the ra pid development of computer technology, computer autonomous control syst ems have played a n important role in indus trial, military, medical and ot her fields.Howeve r, be cause of the traditional PC machine of large volume, inconvenient carrying and high cost, the appli cation of ma ny aspects is limited. Embedded system having a good pros pect with mi niatur iza tion, low power consumption, high performance has been used in the smart home, securi ty, services, etc. The embedded autonomous obstacle avoidance pla tform, which belongs to a part of video surveillance, re lying on monocul ar visual perception of the environment is designed to effectively detect obstacle s on the r oad in the working space and t he environment and cal culate t he distance, and can accurately and quic kly avoid obstacles. The main resear ch and work content in this paper ar e as follows:1. The embedded s ystem hardware and system r esource distribution are analyzed in this thesis, and t he ARM9 architecture of the Linux kernel boot and Nandflash bare driver programs are designed wi th cut ting and transplantation of embedded Linux and making the root fil e system, t hen Qt, OpenCV and image libraries are cross-c ompiled and migr ated, at the same time the developme nt environment has been built. Fi nally, the motor driver of the mobile platform is pl anned.2. Obst acle detection and l ocalization algorithms are designe d and implemented by a machine vision. First, the image captured has been converted from RGB space to Hsv space with the ini tia l processing. With H space of the Hsv space divided i nto t hr ee blocks, color moment features of these blocks a r e compared to dete ct obstacles. In order to locate obstacles correctly and measure di stance between obst acl es and the robot for an ima ge in which obstacles are det ect ed, a r ela tionship between the mobile pl atform coordinates and the vision coordinates i s deduced by geometrical reasoning. Then the direction angle of the mobile platform and the obstacle is calculated.3. The fuzzy control algorithm has been used in the de sign of the mobil e pla tform for common obstacle avoidance problems, and fuzzy controller is designed according to the actual e nvironment and the correspondi ng rules are established. Then the output rotation angl e i s obtained by fuz zy controller which can receive t he obstacle distance which has been mea sured and deviation angle to the original direc tion. Finally, the output rotationangl e is submitted to the mot or driver of t he platfor m to complete some output action.4. With the analysis of the overall design of autonomous mobile platform, the design of t he image acquis iti on modul e has been completed. Then this platform is Integrated,and code is debugged and optimiz ed. Through the analysi s of the experimental results and real-time of the system, the results show that the s ystem is real-time and feasible, and this syst em re ali zes the fa st detection, localization with t he image i nformation and meets avoi dance of obstacle through the e stablishe d control system.