基于小型Zynq SoC硬件加速的改进TINY YOLO实时车辆检测算法实现

doi:10.11772/j.issn.1001-9081.2018051134

计算机应用 ›› 2019, Vol. 39 ›› Issue (1): 192-198.DOI: 10.11772/j.issn.1001-9081.2018051134

基于小型Zynq SoC硬件加速的改进TINY YOLO实时车辆检测算法实现

张雲轲, 刘丹

电子科技大学电子科学技术研究院, 成都 611731

收稿日期:2018-06-01 修回日期:2018-06-21 出版日期:2019-01-10 发布日期:2019-01-21
通讯作者: 刘丹
作者简介:张雲轲(1993-),男,四川成都人,硕士研究生,主要研究方向:人工智能、高性能计算、深度神经网络、模式识别;刘丹(1969-),男,四川成都人,副教授,博士,主要研究方向:网络安全、Web数据挖掘、图像处理。

Real-time implementation of improved TINY YOLO vehicle detection algorithm based on Zynq SoC hardware acceleration

ZHANG Yunke, LIU Dan

Research Institute of Electronic Science and Technology, University of Electronic Science and Technology of China, Chengdu Sichuan 611731, China

Received:2018-06-01 Revised:2018-06-21 Online:2019-01-10 Published:2019-01-21

摘要/Abstract

摘要： 针对TINY YOLO车辆检测算法计算量过大，且在小型嵌入式系统中难以达到实时检测要求的问题。利用小型Zynq SoC系统的架构优势以及TINY YOLO的网络权值中存在大量接近零的权值参数这一特点，提出硬件并行加速的改进算法，称为浓缩小型深度网络（Xerantic-TINY YOLO，X-TINY YOLO）车辆检测算法。首先对TINY YOLO中网络结构进行压缩；其次采用高效多级流水线、流水线内全并行的方式对卷积计算部分进行算法加速；最后提出与网络结构相配合的数据切割和传输方案。实验结果表明，X-TINY YOLO仅消耗50%的片内硬件资源，可在相对于GPU和CPU性价比更高更适合嵌入式场景的Zynq SoC系统上实现，且其检测速度达到24帧/s，满足车辆检测的实时性要求。

关键词: 车辆检测, 机器视觉, TINY YOLO, Zynq-7020, 硬件加速

Abstract: TINY YOLO (TINY You Only Look Once) vehicle detection algorithm requires much amount of calculation which makes it difficult to achieve real-time detection in small embedded systems. Because plenty of zero values exist in a network weight matrix which makes the network a sparse structure, an improved version of TINY YOLO vehicle detection algorithm, called Xerantic-TINY YOLO (X-TINY YOLO), was proposed and accelerated in parallel way using architectural advantages of small Zynq SoC system. Original network structure of TINY YOLO was compressed and the operations of convolution steps were accelerated in parallel by using high efficient multistage pipeline. All multiply-add operations were concurrently executed within each stage of pipeline. By matching network structure, a method of data segmentation and transfer was also proposed. The experimental results show that, X-TINY YOLO only consumes 50% hardware resources on chip, and it can be implemented on small Zynq SoC systems which have higher performance-price ratio than GPU and CPU and is suitable for embedded implementation scenes. Its detection speed reaches 24 frames per second, which meets the requirement of real-time vehicle detection.

Key words: vehicle detection, machine vision, TINY You Only Look Once (TINY YOLO), Zynq-7020, hardware acceleration

中图分类号:

张雲轲, 刘丹. 基于小型Zynq SoC硬件加速的改进TINY YOLO实时车辆检测算法实现[J]. 计算机应用, 2019, 39(1): 192-198.

ZHANG Yunke, LIU Dan. Real-time implementation of improved TINY YOLO vehicle detection algorithm based on Zynq SoC hardware acceleration[J]. Journal of Computer Applications, 2019, 39(1): 192-198.

参考文献

[1] 傅沈文. 复杂环境下基于图和条件随机域的运动车辆检测[J].计算机应用,2012,32(6):1581-1584.(FU S W. Moving vehicle detection in complex environments based on graph and conditional random field[J]. Journal of Computer Applications, 2012, 32(6):1581-1584.)
[2] 陈艳,严腾,宋俊芳,等.基于高斯混合模型和AdaBoost的夜间车辆检测[J].计算机应用,2018,38(1):260-263.(CHEN Y, YAN T, SONG J F, et al. Night-time vehicle detection based on Gaussian mixture and AdaBoost[J]. Journal of Computer Applications, 2018, 38(1):260-263.)
[3] 王德宇,徐友春,李永乐,等.基于深度学习的车辆检测方法[J].计算机与现代化,2017(8):56-60.(WANG D Y, XU Y C, LI Y L, et al. Vehicle detection based on deep learning[J]. Computer and Modernization, 2017(8):56-60.)
[4] SRINIVASA N. Vision-based vehicle detection and tracking method for forward collision warning in automobiles[C]//IV 2002:Proceedings of the 2002 IEEE Intelligent Vehicle Symposium. Piscataway, NJ:IEEE, 2002:626-631.
[5] SCHAMM T, VON CARLOWITZ C, ZOLLNER J M. On-road vehicle detection during dusk and at night[C]//IV 2010:Proceedings of the 2010 IEEE Intelligent Vehicles Symposium. Piscataway, NJ:IEEE, 2010:418-423.
[6] CHANG W C, CHO C W. Online boosting for vehicle detection[J]. IEEE Transactions on Systems, Man & Cybernetics Part B, 2010, 40(3):892-902.
[7] REZAEI M, TERAUCHI M. Vehicle detection based on multi-feature clues and Dempster-Shafer fusion theory[C]//PSIVT 2013:Proceedings of the 2013 Pacific-Rim Symposium on Image and Video Technology. Berlin:Springer, 2013:60-72.
[8] REN S, HE K, GIRSHICK R, et al. Faster R-CNN:towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6):1137-1149.
[9] REDMON J, FARHADI A. YOLO9000:better, faster, stronger[C]//CVPR 2017:Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Washington, DC:IEEE Computer Society, 2017:6517-6525.
[10] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[J]. ArXiv Preprint, 2015, 2015:1512.03385.
[11] MARTENS J. Deep learning via Hessian-free optimization[C]//ICML 2010:Proceedings of the 2010 International Conference on International Conference on Machine Learning. Madison, WI:Omnipress, 2010:735-742.
[12] AN D C, MEIER U, MASCI J, et al. Flexible, high performance convolutional neural networks for image classification[C]//IJCAI 2011:Proceedings of the 2011 International Joint Conference on Artificial Intelligence. Menlo Park, CA:AAAI Press, 2011:1237-1242.
[13] CHEN Y, SUN N, TEMAM O, et al. DaDianNao:a machine-learning supercomputer[C]//Proceedings of the 201447th Annual IEEE/ACM International Symposium on Microarchitecture. Piscataway, NJ:IEEE, 2015:609-622.
[14] YUN S B, KIM Y J, DONG S S, et al. Hardware implementation of neural network with expansible and reconfigurable architecture[C]//ICONIP 2002:Proceedings of the 9th International Conference on Neural Information Processing. Piscataway, NJ:IEEE, 2002:970-975.
[15] FARABET C, MARTINI B, CORDA B, et al. NeuFlow:a runtime-reconfigurable dataflow processor for vision[C]//CVPR 2011 Workshops:Proceedings of the 2011 IEEE Conference on Computer Vision and Pattern Recognition. Washington, DC:IEEE Computer Society, 2011:109-116.
[16] SANKARADAS M, JAKKULA V, CADAMBI S, et al. A massively parallel coprocessor for convolutional neural networks[C]//ASAP 2009:Proceedings of the 2009/20th IEEE International Conference on Application-Specific Systems, Architectures and Processors. Piscataway, NJ:IEEE, 2009:53-60.

基于小型Zynq SoC硬件加速的改进TINY YOLO实时车辆检测算法实现

Real-time implementation of improved TINY YOLO vehicle detection algorithm based on Zynq SoC hardware acceleration

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

[1]	王晓峰, 蒋彭龙, 周辉, 赵雄波. 面向卷积神经网络的高并行度FPGA加速器设计[J]. 计算机应用, 2021, 41(3): 812-819.
[2]	张亚洲, 卢先领. 基于改进相干增强扩散与纹理能量测度和高斯混合模型的导光板表面缺陷检测方法[J]. 计算机应用, 2020, 40(5): 1545-1552.
[3]	刘子威, 邓春华, 刘静. 基于非对称沙漏网络结构的目标检测算法[J]. 计算机应用, 2020, 40(12): 3526-3533.
[4]	郜明, 任德均, 胡云起, 付磊, 邱吕. 基于机器视觉和轻量级卷积网络的安瓿瓶包装质量检测算法[J]. 计算机应用, 2020, 40(10): 2899-2903.
[5]	石昌友, 王美丽, 刘欣然, 黄慧丽, 周德强, 邓干然. 基于机器视觉的不同类型甘蔗茎节识别[J]. 计算机应用, 2019, 39(4): 1208-1213.
[6]	徐子豪, 黄伟泉, 王胤. 基于深度学习的监控视频中多类别车辆检测[J]. 计算机应用, 2019, 39(3): 700-705.
[7]	杨继东, 孙兆琦, 王飞龙. 基于视觉抓取的并联桁架机器人最优路径控制[J]. 计算机应用, 2019, 39(3): 913-917.
[8]	刘保成, 朴燕, 唐悦. 基于时空正则化的视频序列中行人的再识别[J]. 计算机应用, 2019, 39(11): 3216-3220.
[9]	范宏伟, 胡宇翔, 兰巨龙. 两段式虚拟网络功能硬件加速资源部署机制[J]. 计算机应用, 2018, 38(9): 2575-2580.
[10]	王林, 张鹤鹤. Faster R-CNN模型在车辆检测中的应用[J]. 计算机应用, 2018, 38(3): 666-670.
[11]	陈艳, 严腾, 宋俊芳, 宋焕生. 基于高斯混合模型和AdaBoost的夜间车辆检测[J]. 计算机应用, 2018, 38(1): 260-263.
[12]	李肖, 葛宝臻, 罗其俊, 李云鹏, 田庆国. 自由双目立体视觉摄像机动态外参数的获取[J]. 计算机应用, 2017, 37(10): 2888-2894.
[13]	赵树言, 路艳雪, 韩晓霞. 基于最近邻域像素梯度的视频背景快速提取[J]. 计算机应用, 2016, 36(8): 2322-2326.
[14]	李杰, 周浩, 张晋, 高赟. 基于粒子群优化的模板匹配跟踪算法[J]. 计算机应用, 2015, 35(9): 2656-2660.
[15]	孟祥扬, 林琦. 基于NetMagic平台的MD5算法硬件加速模型[J]. 计算机应用, 2015, 35(4): 991-995.