Design And Research About Binocular Autonomous Car System

The word "robot" first appeared in 1920, while the first robot come out in 1959. It is the first industrial robot developed by Unimation, the corporation founded by Mr. Engelberger. Today,the robot technology has become a comprehensive discipline called Robotics. It covers a wide range of research and application fields, including the kinematics, dynamics, sensor technology,control technology, etc. This dissertation will introduce the autonomous car, also known as wheeled robot, which is a very important branch of robot research field and has formed a specialized research direction at present. As the evolution of sensors and the raising of the concept of "intelligent", autonomous car is gaining increasingly various intelligent functions. This dissertation will focus on researching the cruise control system, motor control system, binocular vision measurement system, and wireless remote control system of the autonomous car that has the binocular vision photogrammetry function.This dissertation completed the hardware and software design for binocular autonomous car.The hardware platform adopted TMS320F2812 from TI Cooperation as core processor. And it implement the modular design as well, which including autonomous cruise module, binocular distance measurement module, power module, wireless transmission module, drive motor control module and LCD module. The software part designed specific drivers for each hardware module in autonomous car; calibrated the camera in binocular vision photogrammetry;analyzed and simulated the ranging algorithm; built the target calculating model and derived formula of ranging and target 3D coordinate equation, providing the theoretical basis for the binocular vision photogrammetry.For the binocular vision measuring module, open source computer vision library will be used to calibrate binocular camera, and the calibration method based on checkerboard plane, which putting forward by Zhengyou Zhang, will be adopted to calculate the internal and external parameter matrix, translation matrix and rotation matrix. For the stereo calibration, this dissertation will use Bouguest algorithm to implement it.For the visual measuring methods, this dissertation will present an innovative method that combing binocular measurement in active measuring method with structural lighting measurement in passive measuring method. This method will dramatically reduce the difficulty of target extraction, short the time of feature matching, drop the computation load of system and finally,implement the function of vision measurement on TMS320F2812.For the establishment of hardware platform based on embedded system, since the traditionalvision measuring has huge amount of calculation and the system has to achieve other control function in the meantime, it is always need DSP and FPGA serve together. The binocular ranging measurement algorithm in this dissertation, however, is much simper and easier to implement. The most important is that it can be fulfilled on the TMS320F2812, which is an industrial control level DSP. Eventually, It will form the binocular vision ranging module after integrating camera calibration, target location, distance measurement and LCD display into DSP.