摘要
光流法是计算机视觉中一个基础性的算法,可广泛应用于运动检测、运动估计、视频分析等领域。但光流法最大的问题是计算复杂、速度慢,限制了它在实际系统尤其是嵌入式系统中的应用。利用最新的高层综合(HLS)语言与传统的硬件描述语言相结合,在Xilinx的FPGA异构系统芯片(即ZYNQ)平台上,以软硬件协同的工作方式,设计了基于Horn-Schunck稠密光流法的硬件加速器。实验证明,对于640×480大小的图片,软硬件协同处理比纯软件处理的计算性能提高了34倍,执行时间从24.40 s降低到0.71 s。
Techniques of optical flow computation are widely used in many video/image based applications such as motion detection, motion estimation and video analysis etc. However, high-quality optical flow algorithms are computationally intensive. Slow computation limits the applicability of optical flow computation in real-world applications, especially in embedded systems. In this paper, an implementation of Horn-Schunck optical flow algorithm based on Xilinx ZYNQ is presented. The High-Level Synthesis（HLS）language together with traditional hardware description language is used to describe optical flow accelerator in the software-hardware co-processing mode. Taking resolution 640 ＆#215; 480 as instance, the result shows that FPGA-accelerated HS outperforms 34x than the pure software vision on ZYNQ. The execution time is decreased from 24.40 s to 0.71 s.
出处
《计算机工程与应用》
CSCD
2014年第18期44-49,共6页
Computer Engineering and Applications
基金
国家自然科学基金(No.60703106
No.61170121
No.61202312)
关键词
光流加速器
ZYNQ
高层综合语言
软硬件协同处理
可编程器件
optical flow accelerator
ZYNQ
high-level synthesis language
software-hardware co-processing
Field-Programmable Gate Array （FPGA）
