Development Of A Machion Vision Auto-navigation System Based On DSP

Stereo vision on the stereo vision had a huge development in recent years. It has widely applied to the industrial assembly and inspection, medical image analysis, obstacle detection, entertainment, human-computer interaction and many other fields. Automatic navigation technology has become a hot spot of three-dimensional visual application. However, along with the converting of 2D to 3D, image processing system is facing large volumes, high-precision challenge. This thesis aims to automatic navigation application requirement of real-time video image collection, hardware, software and automatic navigation algorithms for a comprehensive research.This paper put forward a design scheme of real-time video image acquisition and processing hardware platform based on DSP firstly. The DSP of this scheme is selected TMS320DM6437 of Davinci series used in digital video development of TI company, It through video camera (CCDC) collect real-time video image and translated into analog video signal, then will be sent to analog video signal to TVP5146 video A/D conversion chip, the digital video signal will be sent to TMS320DM6437 video processing front end (VPFE) by the hardware connection on DM6437 board after conversion, and output it to the video output interface circuit, then video output interface circuit convert video signal into analog signal and output it. The hardware platform uses the mother-child board structure. The mother board is the DSP core board, it divided into video image processing module, data storage module and the connector to the son board, son board mainly including video codec module, video input/output modules and power modules. Signals are simulated for the signal-integrity of the PCB board in this paper. This paper also discusses the characteristics of image tracking tags based on algorithms, the image matching tracking algorithm feature matching method to implement regional matching method than much faster image analysis processing, greatly reducing the automatic tracking reaction time.