目标跟踪算法的并行优化

计算机工程与科学

目标跟踪算法的并行优化

陈伟1,祝恩1,刘天航1,殷建平1,邱明辉2

（1.国防科学技术大学计算机学院，湖南长沙 410073;2.解放军总医院医学信息情报所,北京 100039）

收稿日期:2016-07-10 修回日期:2016-09-09 出版日期:2016-11-25 发布日期:2016-11-25
基金资助:
国家自然科学基金（61170287,61232016）

Parallel optimization of target tracking algorithms

CHEN Wei1，ZHU En1，LIU Tianhang1，YIN Jianping1，QIU Minghui2

（1.College of Computer,National University of Defense Technology,Changsha 410073;

2.Medical Informatics Institute of Chinese PLA General Hospital,Beijing 100039,China）

Received:2016-07-10 Revised:2016-09-09 Online:2016-11-25 Published:2016-11-25

摘要/Abstract

摘要：

目标跟踪是计算机视觉领域一个重要的研究方向，近年来学者提出了众多优秀的目标跟踪算法，但许多算法的低实时性制约了其在应用场景中的有效性。针对这些算法，提出了一个通用的跟踪模型，并针对此模型提出了一个可行的并行优化方案。之后使用SCM算法验证了所提出的并行优化方案。在四核CPU的环境下，并行后的SCM算法相比于未并行的算法取得了348倍的并行加速比，并且比原算法Matlab+C程序的运行速度快了约30倍,这说明了所提出的并行优化方案的有效性。

关键词: 目标跟踪, 跟踪模型, 并行优化

Abstract:

Object tracking is an important research area in computer vision. Researchers have proposed a number of excellent object tracking algorithms in recent years. However, the poor realtime performance of these algorithms restricts its effectiveness in application scenarios. Based on these algorithms, we design a general tracking model and propose a feasible parallel optimization scheme for the model. We also use the sparsitybased collaborative model (SCM) algorithm to validate the proposed scheme. In a fourcore CPU environment, the parallel algorithm achieves a speedup of 3.48 times compared with the sequential algorithm, and it is about 30 times faster than the MATLAB+C program of the original algorithms, thus verifying the effectiveness of the proposed parallel optimization scheme.

Key words: object tracking, tracking model, parallel optimization

陈伟1,祝恩1,刘天航1,殷建平1,邱明辉2. 目标跟踪算法的并行优化[J]. 计算机工程与科学.

CHEN Wei1，ZHU En1，LIU Tianhang1，YIN Jianping1，QIU Minghui2. Parallel optimization of target tracking algorithms[J]. Computer Engineering & Science.

参考文献

［1］Wei Z,Huchuan L,MingHsuan Y.Robust object tracking via sparse collaborative appearance model［J］.IEEE Transactions on Image Processing,2014,23（5）:23562368．

［2］Zhong W.Robust object tracking via sparsitybased collaborative model［C］∥Proc of 2012 IEEE Conference on Computer Vision and Pattern Recognition,2012:18381845.

［3］Hare S, Saffari A, Torr P H S. Struck:Structured output tracking with kernels［C］∥Proc of the 2011 International Conference on Computer Vision,2011:263270．

［4］Li Y,Zhu J,Hoi S C H.Reliable patch trackers:Robust visual tracking by exploiting reliable patches［C］∥Proc of 2015 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）,2015:353361．

［5］Wu Y,Lim J,Yang M H.Online object tracking:A benchmark［C］∥
2013 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）,2013:24112418．

［6］Zhao W,Ma H,He Q.Parallel Kmeans clustering based on MapReduce［C］∥Proc of the 1st International Conference on Cloud Computing,2009:674679．

［7］Dean J,Ghemawat S.MapReduce:Simplified data processing on large clusters.［J］. Proceedings of Operating Systems Design and Implementation （OSDI),2004,51（1）:107113．

［8］Cai Yang,Li GuangYao,Wang Hu.Research and implementation of parallel particle swarm optimization based on CUDA［J］.Application Research of Computers,2013,30（8）:24152418.（in Chinese）

［9］Nvidia B.CUDA programming guide,231 edition［C］//http://developer.download.nvidia.com/compute/cuda/2_3/toolkit/docs/ NVIDIA_CUDA_Programming_ Guide_23.pdf.2010．

［10］Jasika N, Alispahic N, Elma A, et al. Dijkstra’s shortest path algorithm serial and parallel execution performance analysis［C］∥
Proc of the 35th International Convention on MIPRO,2012:18111815．

［11］Cui Y,Liao X,Zhu X,et al.BMIC:An ultrafast threelevel parallel sequence aligner using MIC［J］.Interdisciplinary Sciences Computational Life Sciences,2015,8（1）:17．

［12］Su H,Wen M,Wu N,et al.Efficient parallel video processing techniques on GPU:From framework to implementation.［J］.The Scientific World Journal,2014,2014（1）:716020716020．

［13］Bolme D S,Beveridge J R,Draper B A,et al.Visual object tracking using adaptive correlation filters［C］∥
Proc of 2010 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）,2010:25442550．

［14］Henriques J F,Caseiro R,Martins P,et al.Exploiting the circulant structure of trackingbydetection with kernels［M］∥Computer VisionECCV 2012.Berlin:Springer Berlin Heidelberg,2012:702715．

［15］Henriques J F, Caseiro R, Martins P, et al.Highspeed tracking with kernelized correlation filters［J］.IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37（3）:583596．

［16］Xu Tingfa,Zhao Sihong,Zhou Shengbing,et al.Particle swarm optimizer tracking based on DSP parallel system［J］.Optics & Precision Engineering,2009,17（9）:22362240.（in Chinese）

［17］Cao Ran.Object tracking using multiple instance learning and parallel application［D］.Nanjing:Nanjing Normal University,2014.（in Chinese）

［18］Pan Liang,Guo Gaizhi,Song Xinmeng.A research on parallel algorithm of matrix multiplication based on OpenMP［J］.Journal of Baoji University of Arts & Sciences,2014,34(1):2123.（in Chinese）

［19］Dong J X,Krzyzak A,Suen C.A fast parallel optimization for training support vector machine［C］∥Proc of International Conference on Machine Learning & Data Mining in Pattern Recognition,2003:96105．

［20］Wang Wei. The research on loop vectorization method in MATLAB programming［J］.Journal of Changchun University,2010,20(1):5759.（in Chinese）

附中文参考文献：

［8］蔡勇,李光耀,王琥.基于CUDA的并行粒子群优化算法的设计与实现［J］.计算机应用研究,2013,30（8）:24152418．

［16］许廷发,赵思宏,周生兵，等.DSP并行系统的并行粒子群优化目标跟踪［J］.光学精密工程,2009,17（9）:22362240．

［17］曹冉.基于多示例学习的目标跟踪算法及其并行化研究［D］.南京：南京师范大学,2014．

［18］潘亮,郭改枝,宋鑫梦.基于OpenMP矩阵相乘并行算法的设计［J］.宝鸡文理学院学报（自然科学版）,2014,34（1）:2123．

［20］王薇.MATLAB的循环向量化编程方法研究［J］.长春大学学报（自然科学版）,2010,20（1）:5759．

[1]	崔克彬, 崔叶微. 基于卷积和Transformer的断路器动触头跟踪方法研究[J]. 计算机工程与科学, 2023, 45(07): 1236-1244.
[2]	杨冬英, 贺江鹏 . 针对机动目标的改进型交互多模型跟踪算法研究[J]. 计算机工程与科学, 2022, 44(04): 686-691.
[3]	马现, 王勇献, 朱小谦, 屠厚旺, 李朋, 颜恺壮. 谱方法求解水声传播问题的优化与并行[J]. 计算机工程与科学, 2022, 44(03): 381-389.
[4]	冯琪堯, 张惊雷, . 基于混合注意力机制的目标跟踪算法[J]. 计算机工程与科学, 2022, 44(02): 276-282.
[5]	方梦华, 姜添, . 基于无监督学习的无人机目标跟踪[J]. 计算机工程与科学, 2021, 43(06): 1024-1031.
[6]	陈云, 王梦园, 柴晓楠, 商建东, . 面向FT-M7002的高斯滤波算法优化实现[J]. 计算机工程与科学, 2021, 43(05): 799-806.
[7]	张晓丽, 张龙信, 肖满生, 左国才. 快速多域卷积神经网络和光流法融合的目标跟踪[J]. 计算机工程与科学, 2020, 42(12): 2217-2222.
[8]	徐亮, 张江, 张晶, 杨亚琦. 基于VGG网络的鲁棒目标跟踪算法[J]. 计算机工程与科学, 2020, 42(08): 1406-1413.
[9]	李星辰，柳晓鸣，成晓男. 融合YOLO检测的多目标跟踪算法[J]. 计算机工程与科学, 2020, 42(04): 665-672.
[10]	黄云明1，张晶1,2，喻小惠1，陶涛3，龚力波4. 改进STC和SURF特征联合优化的目标跟踪算法[J]. 计算机工程与科学, 2019, 41(10): 1795-1802.
[11]	火元莲，曹鹏飞，董俊松，石明. 基于融合特征的多尺度快速相关滤波跟踪算法[J]. 计算机工程与科学, 2019, 41(03): 559-566.
[12]	张晶，王旭,范洪博. 自适应目标变化的时空上下文抗遮挡跟踪算法[J]. 计算机工程与科学, 2018, 40(09): 1653-1661.
[13]	张丹1，陈兴文1，赵姝颖2. 基于水平集的TLD目标跟踪改进算法[J]. 计算机工程与科学, 2017, 39(05): 984-991.
[14]	李娜1,2,3，赵祥模1，赵凤2,3，刘卫华2,3，王倩2,3. 基于外观模型的目标跟踪算法研究进展[J]. 计算机工程与科学, 2017, 39(03): 524-533.
[15]	陈金广，江梦茜，马丽丽. 部分监测区域重叠条件下的扩展目标跟踪[J]. J4, 2016, 38(06): 1231-1237.