基于FPGA+DSP的智能车全景视觉系统  

Omni-vision system of intelligent car based on FPGA&DSP

作　　者：周渝斌[1] 

机构地区：[1]上海工程技术大学机械工程学院,上海201620

出　　处：《电子技术应用》2011年第3期38-41,共4页Application of Electronic Technique

基　　金：上海高校选拔培养优秀青年教师科研专项基金(GJD08017)

摘　　要：为实现智能车全景视觉系统的应用研究平台,设计了一种基于FPGA+双DSP的实时6通道数字图像采集与处理系统。该系统由两片FPGA与两个DSP组成。第一个FPGA进行多通道视觉图像采集的同步控制、逻辑处理,第二片FPGA辅助DSP进行海量图像数据的高速并行处理。两个ZBTSRAM芯片作为数据输入和输出的高速缓存,每通道的A/D输出与ZBT SRAM接口间进行数位拼接。系统工作时,DSP通过EMIF与FPGA进行高速数据通信,而两个DSP之间通过McBSP进行数据通信。系统工作时使用μC/OS操作系统进行多任务负载均衡管理,最终实现对两路视频信号同时实时采集和处理。In order to realize omni-vision system of intelligent car for auto wandering in real-time processing, an image processing system with 6 vision channels based on FPGA＆DSP is designed. In the system, two ZBT SRAM chips are used as the input and output cache for high data transferring. An FPGA chip is responsible for the core logic controlling and video synchronous. Digital videos are sent to processing module by camlink bus. Data are exchanged by EMIF and McBSP between FPGA and DSPs. EDMA is used for data transferring between SRAM in FPGA and ZBT SRAM. The QDMA is used for 2D data transferring to 1D for face data processing into DSP cache. Tasks are assigned to chips by μC/OS on master DSP. All this together, real-time data sampling and processing for muhi-channel vision is realized.

关 键 词：全景视觉 多通道视觉 FPGA DSP 智能车 

分 类 号：TN911.73[电子电信—通信与信息系统]