The Design And Implementation Of Vehicle Vision Navigation System Based On DSP

With the development of our society and the progress of economy, science and technology is developing as time pasts. Playing an essential and indispensible role in traffic, transportation, and outer space explore in our daily life, technology of ground vehicle is progressing as well and unmanned ground vehicle is the trends of future ground vehicle. Since the unmanned ground vehicle can modify driving direction and speed according to the situation in front of the vehicle automatically, it can operate itself without human being interfering, thus unmanned ground vehicle can avoid accidents caused by driver’s error, and avoid casualties in battlefield. According to the reasons mentioned above, the research of unmanned ground vehicle is meaningful.This paper describes the design of hardware platform. The platform is composed of a high speed DSP of ADI, two high ability CMOS image sensors, a high speed low power CPLD, FLASH and SRRAM. The purpose of hardware is to acquire and process images.The paper describes the principle of binocular stereo vision and designs a calibration method. Based on principle of triangle measure, binocular stereo vision is able to measure the three dimension position in the scene in front of the camera. When measuring, the disparity of the image has to be calculated precisely, and how to calculate disparity is emphasized in this paper. The paper researches the mathematic model of camera, and introduces the Zhang’s calibration method, then shows the result of calibration.The paper then proposes a real-time matching algorithm that fits the DSP platform. The algorithm at first applies DoG filter to the image pair to reduce the high frequent noisy and eliminate the DC composition. Then adopts SAD algorithm that is efficient and effective to compute match cost, after accumulating cost, this algorithm adopts WTA to judge the disparity of two images. Then the algorithm use a series of method to eliminate the false matching point, then a reliable disparity map can be achieved to navigate.The implementation of algorithm on DSP hardware platform is described in detail. After analyzing the disadvantages of C language, this paper introduces the optimization of program written by ASM code. Then the paper analyzes the improvement of ASM code compared with C language. Finally the paper gives the result of implementation of algorithm.