基于低秩表示动态更新投影的在线运动目标检测

计算机工程与科学

基于低秩表示动态更新投影的在线运动目标检测

杨国亮，丰义琴，唐俊，谢乃俊

（江西理工大学电气工程与自动化学院，江西赣州 341000）

收稿日期:2015-08-17 修回日期:2015-10-29 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（51365017，61305019）；江西省科技厅青年科学基金（20132bab211032）

Dynamically updating projection via low rank

representation for online moving objects detection

YANG Guoliang,FENG Yiqin,TANG Jun,XIE Naijun

（School of Electrical Engineering and Automation,Jiangxi University of

Science and Technology,Ganzhou 341000,China）

Received:2015-08-17 Revised:2015-10-29 Online:2016-11-25 Published:2016-11-25

摘要/Abstract

摘要：

视频图像中运动目标检测是机器视觉领域的重要研究内容，旨在将序列图像中的背景和前景进行有效分离。在研究几种典型运动目标检测算法的基础上，提出了一种基于低秩表示动态更新投影的在线运动目标检测算法。采用低秩表示方法对若干连续视频帧进行低秩分解，并将分解所获得的低秩部分对应的左奇异值矩阵的正交补引为投影矩阵；再构建投影模型，拟合出数据的稀疏前景；最后采用视频分段分析法则对投影矩阵进行动态更新，从而保证所分离的背景以及前景的有效性。在Curtain等多个视频数据库上与其他算法进行了对比实验，实验结果表明所提算法具有很好的检测效果，对复杂的运动前景和动态背景的处理表现出很强的鲁棒性。

关键词: 低秩表示, 投影矩阵, 在线运动目标检测

Abstract:

Moving objects detection for video processing,which focuses on dividing video images into foreground and background,is important in computer vision.We analyze several traditional moving detection methods,and propose a dynamically updating projection method for online moving objects detection.The proposed method uses the low rank representation (LRR) method to obtain the low rank part of several continuous video images,and thus the projection can be constructed with orthogonal complement of the left singular matrix from the obtained low rank part.The sparse foreground can be obtained by solving the projection model.Besides,the video can be divided into several uniformlyspaced parts,based on which the projection can be dynamically updated.Experimental results on several video databases such as the Curtain demonstrate that our method has better detection performance than the other methods,and it has strong robustness especially when dealing with dynamic background and complex foreground.

Key words: low rank representation, projection matrix, online moving objects detection

杨国亮，丰义琴，唐俊，谢乃俊. 基于低秩表示动态更新投影的在线运动目标检测[J]. 计算机工程与科学.

YANG Guoliang,FENG Yiqin,TANG Jun,XIE Naijun.

Dynamically updating projection via low rank

representation for online moving objects detection

[J]. Computer Engineering & Science.

参考文献

［1］Papenberg N,Bruhn A,Brox T.Highly accurate optic flow computation with theoretically

justified warping ［J］.International Journal of Computer Vision,2006,67(2):141158.

［2］Stauffer C,Grimson W E L.Adaptive background mixture models for realtime tracking［C

］∥Proc of IEEE Conference on Computer Vision and Pattern Recognition,1999:246252.

［3］Elgammal A,Harwood D,Davis L S.Nonparametric model for background subtraction［C］∥

Proc of the 6th European Conference on Computer Vision,2000:751767.

［4］Barnich O, Droogenbroeck M V. Vibe:A universal background subtraction algorithm for

video sequences［J］.IEEE Transactions on Image Processing,2011,20(6):17091724.

［5］Wright J,Ganesh A,Rao S,et al.Robust principal component analysis:Exact recovery of

corrupted lowrank matrices by convex optimization［C］∥Proc of Advances in Neural

Information Processing Systems (NIPS),2009:20802088.

［6］Candes E,Li Xiaodong,Ma Yi,et al.Robust principal com

ponent analysis［J］.Journal of the ACM,2011,58(3):11.

［7］Qiu C,Vaswani N.Automated recursive projected cs (ReProCS) for realtime video

layering［C］∥Proc of International Conference on Computer Vision and Pattern Recognition

(CVPR),2012:1.

［8］Lin Z,Liu R,Su Z.Linearized alternating direction method with adaptive penalty for low

rank representation,［C］∥Proc of Advances in Neural Information Processing

Systems,2011:612620.

［9］Shi Jiarong,Zheng Xiuyun,Wei Zongtian,et al.Survey on algorithms of lowrank

representation［J］. Application Research of Computers,2013,30(6):16011605.(in Chinese)

［10］Zhuang L,Gao H,Lin Z,et al.Nonnegative low rank and sparse graph for semi

supervised learning,［C］∥Proc of IEEE Conference on Computer Vision & Pattern

Recognition,2012:23282335.

［11］Tropp J A，Gilbert A C.Signal recovery from partial information via orthogonal

matching pursuit［J］.IEEE Transactions on Information Theory，2007，53(12):46554665.

［12］Candes E,Li Xiaodong,Ma Yi,et al.Robust principal component analysis［J］.Journal

of the ACM,2011，58(3):11.

［13］Toyama K, Krumm J, Brumitt B, et al. Wallflower:Principles and practice of

background maintenance［C］∥Proc of IEEE International Conference on Computer

Vision,1999:255261.

［14］Li L,Huang W,Gu I Y H,et al.Statistical modeling of complex backgrounds for

foreground object detection［J］.IEEE Transactions on,Image Processing,2004,13(11):1459

1472.

［15］Prati A,Mikic I,Trivedi M M,et al.Detecting moving shadows:Algorithms and evaluation

［J］.IEEE Transactions on Pattern Analysis and Machine Intelligence,2003,25(7):918923.

［16］De La Torre F,Black M J.A framework for robust subspace learning［J］.International

Journal of Computer Vision,2003,54(13):117142.

［17］Candès E J,Li X,Ma Y,et al.Robust principal component analysis［J］.Journal of the

ACM (JACM),2011,58(3):11.

附中文参考文献：

［9］史加荣,郑秀云,魏宗田,等.低秩矩阵恢复算法综述［J］.计算机应用研究,2013,30(6):1601

1605.

[1]	孙雨浩, 陶洋, 胡昊. 基于判别低秩矩阵恢复和协同表示的遮挡人脸识别[J]. 计算机工程与科学, 2020, 42(12): 2199-2207.
[2]	刘作军，高尚兵. 基于结构化低秩表示和低秩投影的人脸识别算法[J]. 计算机工程与科学, 2018, 40(01): 108-115.
[3]	杨国亮，丰义琴，鲁海荣. 基于低秩表示中稀疏误差的可变光照和局部遮挡人脸识别[J]. J4, 2015, 37(09): 1742-1749.
[4]	杨国亮，罗璐，丰义琴，梁礼明. 基于低秩稀疏图的结构保持投影算法[J]. J4, 2015, 37(08): 1584-1590.
[5]	杨国亮，谢乃俊，王艳芳，梁礼明. 基于低秩稀疏评分的非监督特征选择[J]. J4, 2015, 37(04): 649-656.