一种图像缩放算法的SoC协同加速设计方法

王鹏; 曹云峰; 许蕾; 丁萌; 张洲宇; 曲金秋

doi:10.13700/j.bh.1001-5965.2018.0313

一种图像缩放算法的SoC协同加速设计方法

doi: 10.13700/j.bh.1001-5965.2018.0313

王鹏¹,
曹云峰^1, ,,
许蕾¹,
丁萌²,
张洲宇¹,
曲金秋¹

1.
南京航空航天大学航天学院, 南京 210016
2.
南京航空航天大学民航学院, 南京 211106

基金项目:

国家自然科学基金 61673211

南京航空航天大学博士学位论文创新与创优基金 BCXJ18-11

中央高校基本科研业务费专项资金 kfjj20171502

详细信息

作者简介:
王鹏  男, 硕士研究生。主要研究方向:嵌入式、计算机视觉、无人机先进控制

曹云峰  男, 硕士, 教授, 博士生导师。主要研究方向:无人机飞行控制与导航、机器视觉与人工智能、基于模型的系统设计技术

许蕾  男, 硕士, 讲师。主要研究方向:嵌入式、无人机智能控制

丁萌  男, 博士, 副教授, 硕士生导师。主要研究方向:计算机视觉、无人机导航、制导与控制

通讯作者:
曹云峰, E-mail: cyfac@nuaa.edu.cn

中图分类号: TP391
计量
- 文章访问数: 755
- HTML全文浏览量: 181
- PDF下载量: 394
- 被引次数: 0
出版历程
- 收稿日期: 2018-05-30
- 录用日期: 2018-10-15
- 网络出版日期: 2019-02-20

SoC collaborative acceleration design method for image scaling algorithm

1.
College of Astronautics, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China
2.
College of Civil Aviation, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China

Funds:

National Natural Science Foundation of China 61673211

Fundation of Innovation and Excellence Fund for Doctoral Dissertations in NUAA BCXJ18-11

the Fundamental Research Funds for the Central Universities kfjj20171502

More Information

Corresponding author: CAO Yunfeng, E-mail: cyfac@nuaa.edu.cn

摘要

摘要:
针对无人机自主着陆的跑道检测、识别、跟踪等视觉算法中需要对大量图像进行缩放处理以便后续计算，但又对实时性要求比较高的情况，根据输入输出像素点的映射关系提出了一种适用于硬件加速的图像缩放算法，简化算法结构的同时利用现场可编程门阵列进行模块硬件功能的设计对算法加速，并采用软硬件协同的体系结构搭建实时图像处理系统。实验结果表明，该缩放算法处理精度高、耗时少，且用硬件逻辑实现后，可以进一步提速171倍，硬化后的系统可以通过摄像头获取图像数据，实时处理后在显示器中显示，达到30帧/s的处理速度，可以应用于实时性要求较高的图像处理算法中。
- 图像缩放 /
- 实时图像处理 /
- 片上系统(SoC) /
- 软硬件协同 /
- 无人机视觉
Abstract:
Aimed at the problem that a large number of images need to be scaled in the visual algorithm for the runway detection, recognition and tracking of the unmanned aerial vehicle with high real-time requirement, a new image scaling algorithm suitable for hardware acceleration is proposed based on the mapping relation of the input-output pixel. By simplifying the algorithm structure and using the field programmable gate array to design the hardware function of the module, the algorithm accelerates, and the real-time image processing system is built by the software and hardware cooperative architecture. The experimental results show that the improved scaling algorithm has high precision and less time consumption, and it can speed up by 171 times with the hardware logic. The hardened system can get the image data through the camera, and the real-time processing is displayed in the monitor, which has 30 frame/s processing speed. It can be applied to the image processing algorithm with high real-time requirement.
- image scaling /
- real-time image processing /
- system on chip (SoC) /
- software and hardware collaboration /
- unmanned aerial vehicle vision

HTML全文

图 1 系统硬件结构示意图

Figure 1. Schematic diagram of system hardware structure

下载: 全尺寸图片幻灯片

图 2 求取输出点示意图

Figure 2. Schematic diagram of output point solving

下载: 全尺寸图片幻灯片

图 3 不同算法处理效果图

Figure 3. Processing effect diagram of different algorithms

下载: 全尺寸图片幻灯片

图 4 图像缩放模块仿真波形

Figure 4. Simulation waveform of image scaling module

下载: 全尺寸图片幻灯片

图 5 系统部分模块图

Figure 5. Part of module diagram of system

下载: 全尺寸图片幻灯片

图 6 系统测试图

Figure 6. System test chart

下载: 全尺寸图片幻灯片

图 7 MATLAB和SoC处理后直方图对比

Figure 7. Histogram contrast after MATLAB and SoC processing

下载: 全尺寸图片幻灯片

表 1 定量分析对比

Table 1. Comparison of quantitative analysis

客观评价指标	最邻近插值	双线性插值	双三次插值	本文算法
left	0.0274	0.0268	0.0257	0.0252
MSE	0.0026	0.0021	0.0020	0.0020
PSNR	25.9162	26.7568	27.0400	27.0521
SNR	18.0160	18.8565	19.1397	19.1518

下载: 导出CSV

表 2 4种算法的MATLAB运行时间对比

Table 2. Comparison of MATLAB running time among four algorithms

算法	运行时间/ms
最邻近插值	3.02
双线性插值	13.73
双三次插值	25.66
本文算法	3.44

下载: 导出CSV

表 3 本文算法的MATLAB和SoC处理时间对比

Table 3. Comparison of MATLAB and SoC processing time of proposed algorithm

处理方式	处理时间/ms
MATLAB	3.44
SoC	0.02

下载: 导出CSV

参考文献(16)

[1]	RETTKOWSKI J, BOUTROS A, GOHRINGER D.HW/SW co-design of the HOG algorithm on a Xilinx Zynq SoC[J]. Journal of Parallel and Distributed Computing, 2017, 109(1):50-62. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=893c1fba8b2068f9cb7fc078d4303e3d
[2]	KRYJAK T, KOMORKIEWICZ M, GORGON M.Real-time hardware-software embedded vision system for its smart camera implemented in Zynq SoC[J]. Journal of Real-Time Image Processing, 2016, 12(4):1-37. doi: 10.1007/s11554-016-0588-9
[3]	SENOUCI B, CHARFI I, HEYRMAN B, et al.Fast prototyping of a SoC-based smart-camera:A real-time fall detection case study[J]. Journal of Real-Time Image Processing, 2016, 12(4):649-662. doi: 10.1007/s11554-014-0456-4
[4]	SVETEK A, BLAKE M, HERMIDA M C, et al.The calorimeter trigger processor card:The next generation of high speed algorithmic data processing at CMS[J]. Journal of Instrumentation, 2016, 11(2):201-210. http://adsabs.harvard.edu/abs/2016JInst..11C2011S
[5]	ZARANDY A, NEMETH M, NAGY Z, et al.A real-time multi-camera vision system for UAV collision warning and navigation[J]. Journal of Real-Time Image Processing, 2016, 12(4):709-724. doi: 10.1007/s11554-014-0449-3
[6]	刘镇弢, 李涛, 黄虎才, 等.一种用于实时图像处理的众核结构设计[J].西安电子科技大学学报, 2015, 42(2):95-101. doi: 10.3969/j.issn.1001-2400.2015.02.016 LIU Z T, LI T, HUANG H C, et al.A design of the core structure for real-time image processing[J]. Journal of Xidian University, 2015, 42(2):95-101(in Chinese). doi: 10.3969/j.issn.1001-2400.2015.02.016
[7]	杨帆, 张皓, 马新文, 等.基于FPGA的图像处理系统[J].华中科技大学学报(自然科学版), 2015, 43(2):119-123. http://d.old.wanfangdata.com.cn/Thesis/Y1172118 YANG F, ZHANG H, MA X W, et al.Image processing system based on FPGA[J]. Journal of Huazhong University of Science and Technology(Natural Science Edition), 2015, 43(2):119-123(in Chinese). http://d.old.wanfangdata.com.cn/Thesis/Y1172118
[8]	ZHAI X, ALI A S, AMIRA A, et al.MLP neural network based gas classification system on Zynq SoC[J]. IEEE Access, 2016, 4(2):8138-8146. http://ieeexplore.ieee.org/document/7605493/
[9]	HAN Y, VIRUPAKSHAPPA K, VITORSILVAPINTO E, et al.Hardware/software co-design of a traffic sign recognition system using Zynq FPGAs[J]. Electronics, 2015, 4(4):1062-1089. doi: 10.3390/electronics4041062
[10]	KELLY C, SIDDIQUI F M, BARDAK B, et al.FPGA soft-core processors, compiler and hardware optimizations validated using HOG[C]//International Symposium on Applied Reconfigurable Computing.Berlin: Springer, 2016, 1: 78-90.
[11]	ALTUNCU M A, GUVEN T, BECERIKLI Y, et al.Real-time system implementation for image processing with hardware/software co-design on the Xilinx Zynq platform[J]. International Journal of Information and Electronics Engineering, 2015, 5(6):473-477. doi: 10.7763/IJIEE.2015.V5.582
[12]	KRAJNIK T, SVAB J, PEDRE S, et al.FPGA-based module for SURF extraction[J]. Machine Vision and Applications, 2014, 25(3):787-800. doi: 10.1007/s00138-014-0599-0
[13]	GAO F, HUANG Z, WANG S, et al.Optimized parallel implementation of face detection based on embedded heterogeneous many-core architecture[J]. International Journal of Pattern Recognition and Artificial Intelligence, 2017, 31(7):175-180. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=10.1142_S0218001417560110
[14]	ISHIKAWA S, TAKAHASHI T, WATANABE S, et al.High-speed X-ray imaging spectroscopy system with Zynq SoC for solar observations[J]. Nuclear Instruments and Methods in Physics Research Section A:Accelerators, Spectrometers, Detectors and Associated Equipment, 2017, 22(1):40-52. http://arxiv.org/abs/1711.04372
[15]	CAI W, XU Z, LI Z.A high performance surf image feature detecting system based on Zynq[J]. DEStech Transactions on Computer Science and Engineering, 2017, 11(2):101-110.
[16]	王博.数字图像缩放及其质量评价方法研究[D].哈尔滨: 哈尔滨工程大学, 2015: 97-105. http://cdmd.cnki.com.cn/Article/CDMD-10217-1017245829.htm WANG B.Digital image zoom and its quality evaluation method[D]. Harbin: Harbin Engineering University, 2015: 97-105(in Chinese). http://cdmd.cnki.com.cn/Article/CDMD-10217-1017245829.htm