Research On Binocular Obstacle Avoidance Technology Based On Model Based Design

At present, the technology of the intelligent obstacle avoidance based on binocular vision is one of the promising research areas. Real-time embedded platform used in the development of binocular vision technology mainly include: ARM, FPGA and DSP, etc., but in the traditional embedded development method, hardware and software are independent, it is hard to join it up, and the development cycle is long. Model-based design provide a more rapid and efficient method of embedded system development. This paper mainly studies binocular obstacle avoidance algorithm with model-based design.The outline of this paper is organized as follows: 1. Aimming at the problem of lacking special binocular vision platform for use in the market, TMS320C6416 T DSP was selected as the core processor, and the implementing method of the dual camera capturing images synchronously and storage space allocation was presented. Binocular research platform was constructed at last. 2. In the framework of binocular platform, using underlying software support module of DSP circuit system realized the interaction the hardware and software in SIMULINK.And by adding the underlying implementation of binocular image acquisition libraries, the paper designed and realized model-based design platform for binocular research.3. Camera calibration is a necessary step in 3D computer vision. In order to realize target localization in farther distance and increase accuracy of the measurement, article calibrates the binocular system using Matlab calibration toolbox and traditional calibration method, and the two methods were compared and analysed at last.4. Ranging algorighm based on binocular vision was studied. Matching is the most time-consuming part in 3D range algrithm, in order to improve the real-time performance, this paper processed bi-directional matching in a limited parallax, and based on geometric constraints to eliminate error matching corners. At last, the binocular obstacle avoidance algorithm for simulation experiment was carried out.Finally, the whole platform was tested. The function of binocular distance measurement and obstacle avoidance was realized with model-based design.