张强, 孙静, 王威廉, 康立富. 基于ZYNQ硬件加速OpenCV实时高清显示系统设计[J]. 云南大学学报(自然科学版), 2019, 41(5): 918-924. doi: 10.7540/j.ynu.20190104
引用本文: 张强, 孙静, 王威廉, 康立富. 基于ZYNQ硬件加速OpenCV实时高清显示系统设计[J]. 云南大学学报(自然科学版), 2019, 41(5): 918-924. doi: 10.7540/j.ynu.20190104
ZHANG Qiang, SUN Jing, WANG Wei-lian, KANG Li-fu. A design of hardware accelerated OpenCV Real-Time HD display system based on ZYNQ[J]. Journal of Yunnan University: Natural Sciences Edition, 2019, 41(5): 918-924. DOI: 10.7540/j.ynu.20190104
Citation: ZHANG Qiang, SUN Jing, WANG Wei-lian, KANG Li-fu. A design of hardware accelerated OpenCV Real-Time HD display system based on ZYNQ[J]. Journal of Yunnan University: Natural Sciences Edition, 2019, 41(5): 918-924. DOI: 10.7540/j.ynu.20190104

基于ZYNQ硬件加速OpenCV实时高清显示系统设计

A design of hardware accelerated OpenCV Real-Time HD display system based on ZYNQ

  • 摘要: OpenCV是一个用于PC端图像处理、分析方面的开源函数库.为解决OpenCV对PC端资源依赖程度高、耗时长等问题,研究按照Vivado HLS规范,将C++编写的OpenCV程序封装成Verilog IP核,并导入ZYNQ的PL中;再结合Xilinx官方提供的IP核库,以及通过ADI的LCD控制器-ADV7511,实现了基于Xilinx APSOC平台-ZYNQ,实时硬件加速OpenCV图像处理的高清显示系统设计. 实验结果表明:对OpenCV中的Rgb2Gray、Sobel、Threshold等算法实现了硬件加速,处理一幅图像共计耗时11.12 ms,各硬件模块内部资源占用率为1%至3%;而传统PC端运行OpenCV相应算法、处理同一图像耗时为375 ms. 硬件加速后的高清显示系统能通过HDMI接口清晰、低延时地显示高清图像.

     

    Abstract: OpenCV is a function library of open source for image processing and analysis at PC-side. In order to solve the problem of high dependence on the resources of PC and long time-consuming, in this work, the OpenCV was programmed in C++ and embedded into Verilog IP core according to Vivado HLS specification. Then it was loaded into PL of ZYNQ. By combining it with Xilinx official IP core library and using ADI's LCD controller - ADV7511, it was realized that a design of real-time OpenCV image HD display system accelerated by hardware, based on Xilinx APSOC platform - ZYNQ. The experimental results showed that the time consumption was 11.12 ms for processing a frame of picture using Rgb2Gray, Sobel, Threshold and other algorithms in OpenCV accelerated by hardware. Its occupancy of internal resource for each hardware module just was 1% to 3%. But time consumption was 375 ms for same sting to be done by traditional way running OpenCV at PC. The hardware-accelerated HD display system can display high-definition images with clear and low-latency through the HDMI interface.

     

/

返回文章
返回