基于PCIe高速通信接口的图像处理系统设计  

Design of Image Processing System Based on High Speed Data Transfer Interface of PCIe

作　　者：袁柳 李皓 李勐 涂吉 YUAN Liu;LI Hao;LI Meng;TU Ji(China Academy of Electronics and Information Technology,Beijing 100041,China;Electronic Information School Wuhan University,Wuhan 430072,China)

机构地区：[1]中国电子科技集团公司电子科学研究院,北京100041 [2]武汉大学电子信息学院,武汉430072

出　　处：《科学技术与工程》2019年第22期235-240,共6页Science Technology and Engineering

基　　金：装备预研中国电科联合基金(6141B08080101);国家重点研发计划(2017YFC0821603)资助

摘　　要：面向图像处理数据的高速传输和快速处理需求,设计实现了基于PCIe高速通信接口的图像处理系统.在NetFPGA SUME平台的基础上,借助Riffa PCIe架构实现中央处理器(central processing unit,CPU)和现场可编程门陈列(field-programma-ble gate array,FPGA)高速数据传输,充分发挥PCIe总线接口高效性、灵活性、可扩展、低延迟传输性能.设计统一图像处理和管理硬件接口,支持高效实现卷积神经网络(convolutional neural network,CNN)手写字符识别的FPGA加速处理.测试表明:PCIe传输速度可以达到2.86 GB/s;CNN手写字符识别单张图片运行时间为1.58 ms.研究结果可有效提升图像处理系统的数据传输和处理能力.In order to meet the demand of high-speed data transmitting and high-speed image processing,an image processing system was proposed via high speed PCIe interface.Based on NetFPGA SUME,Riffa architecture was used to obtain PCIe interconnection between the central processing unit(CPU)and the field-programmable gate array(FPGA),which gave full play to the efficiency,adaptability,extendibility,low latency of PCIe interface.The unified interfaces of data accessing and managing were implemented,and the convolutional neural network(CNN)handwritten character recognition was efficiently proposed by FPGA.Experimental results show that the transfer rate can reach 2.86 GB/s;the time of CNN handwritten character recognition can be 1.58 ms.The results can promote the capability of transmission and data processing of image processing system.

关 键 词：PCIE 现场可编程门陈列(FPGA) 图像处理 手写字符识别 卷积神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]