针对运动目标检测算法在传统PC端上实时性较差的问题,设计了一种基于ZYNQ硬件加速的运动目标实时检测系统。将摄像头采集的彩色视频流转换为灰度视频流并进行图像处理来实现运动目标检测,并将检测后的结果与原彩色视频流叠加来显示实时检测结果;选用经典的帧差法,并在ZYNQ平台上设计和实现该算法,在VDMA存储中使用乒乓操作加速,中值滤波进行图像处理时使用流水线操作并行加速,大大地提高了算法处理速度。设计实现后对传统的CPU+OpenCV实现横向对比分析,结果表明ZYNQ平台在实时性上具有明显优势。
Abstract
Aiming at the poor real-time performance of the moving target detection algorithm on the traditional PC side,a real-time detection system for moving targets based on ZYNQ hardware acceleration is designed.In the design,the color video stream collected by the camera is converted into a gray-scale video stream and performed.Image processing is used to achieve moving target detection,and the detected results are superimposed with the original color video stream to display real-time detection results.The classic frame difference method is selected,and the algorithm is designed and implemented on the ZYNQ platform.The ping-pong operation is used to accelerate the VDMA storage, and the pipeline operation is used for parallel acceleration when the median filter is used for image processing,which greatly improves the processing speed of the algorithm.After the design is implemented,the traditional CPU+OpenCV implementation is horizontally compared and analyzed,and the results show that the ZYNQ platform has a huge advantage in real-time performance.
关键词
XC7Z020 /
硬件加速 /
帧差法 /
运动目标检测
Key words
XC7Z020 /
hardware acceleration /
frame-difference method /
moving target detection
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 张军阳,王慧丽,郭阳,等.深度学习相关研究综述[J].计算机应用研究,2018,35(7):1921-1928,1936.
[2] Lu Ziwei,Wu Chengdong,Yu Xiaosheng.Single Image Super Resolution Using Nearest Neighbor Local Gaussian Process Regression[C]//ICMLC 2018:2018 10th International Conference on Machine Learning and Computing,2018.
[3] Shouren Huang,Kenta Shinya,Niklas Bergstrm,et al.Dynamic compensation robot with a new high-speed vision system for flexible manufacturing[J].The International Journal of Advanced Manufacturing Technology,2018,95(9):4523-4533.
[4] Tang J W,Shaikh-Husin N,Sheikh U U,et al.FPGA-Based Real-Time Moving Target Detection System for UnmannedAerial Vehicle Application[J].International Journal of Reconfigurable Computing,2016(3):1.
[5] 华夏,王新晴,王东,等.基于改进SSD的交通大场景多目标检测[J].光学学报,2018,38(12):221-231.
[6] 郭诚欣,陈红,孙辉,等.基于现代硬件的并行内存排序方法综述[J].计算机学报,2017,40(9):2070-2092.
[7] Amir HajiRassouliha.Suitability of recent hardware accelerators (DSPs,FPGAs,and GPUs) for computer vision and image processing algorithms[J].Signal Processing: Image Communication,2018(68):101-119.
[8] Seongseop Kim,Jeonghun Cho,Daejin Park.Moving-Target Position Estimation Using GPU-Based Particle Filter for IoT Sensing Applications[J].Applied Sciences,2017,7(11).
[9] 黄彪,孙馨喆.基于相位相关算法的运动目标速度测量方法的研究[J].国外电子测量技术,2021,40(4):110-113.
[10] 韩悦,金晅宏,郭旭,等.视频序列中的运动目标检测算法研究[J].电子测量技术,2019,42(13):103-106.
[11] 吕慷,张旭秀,李卫东.基于人眼识别原理的运动目标检测方法[J].电子测量技术,2019,42(4):65-69.
[12] 山丹,丛国涛.基于FPGA的动态目标识别与跟踪系统设计[J].电子测量技术,2019,42(10):132-136.
[13] 李炳奇.基于FPGA的目标检测与跟踪[D].北京:北京邮电大学,2019.
[14] 陶勇.运动目标检测跟踪的FPGA系统实现[D].昆明:昆明理工大学,2018.
[15] 邢凯,李彬华,陶勇,等.基于FPGA的运动目标实时检测跟踪算法及其实现技术[J].光学技术,2020,46(2):158-166.
[16] Francesco G B,De Natale,Giulia Boato.Detecting Morphological Filtering of Binary Images[J].IEEE Trans. Information Forensics and Security,2017,12(5):1207-1217.
PDF(1364 KB)
