Two Maximum Entropy-Based Algorithms for Running Quantile Estimation in Nonstationary Data Streams

被引:16
作者
Arandjelovic, Ognjen [1 ]
Pham, Duc-Son [2 ]
Venkatesh, Svetha [1 ]
机构
[1] Deakin Univ, Sch Informat Technol, Geelong, Vic 3125, Australia
[2] Curtin Univ, Dept Comp, Perth, WA 6102, Australia
关键词
Histogram; novelty; surveillance; video;
D O I
10.1109/TCSVT.2014.2376137
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
The need to estimate a particular quantile of a distribution is an important problem that frequently arises in many computer vision and signal processing applications. For example, our work was motivated by the requirements of many semiautomatic surveillance analytics systems that detect abnormalities in close-circuit television footage using statistical models of low-level motion features. In this paper, we specifically address the problem of estimating the running quantile of a data stream when the memory for storing observations is limited. We make the following several major contributions: 1) we highlight the limitations of approaches previously described in the literature that make them unsuitable for nonstationary streams; 2) we describe a novel principle for the utilization of the available storage space; 3) we introduce two novel algorithms that exploit the proposed principle in different ways; and 4) we present a comprehensive evaluation and analysis of the proposed algorithms and the existing methods in the literature on both synthetic data sets and three large real-world streams acquired in the course of operation of an existing commercial surveillance system. Our findings convincingly demonstrate that both of the proposed methods are highly successful and vastly outperform the existing alternatives. We show that the better of the two algorithms (data-aligned histogram) exhibits far superior performance in comparison with the previously described methods, achieving more than 10 times lower estimate errors on real-world data, even when its available working memory is an order of magnitude smaller.
引用
收藏
页码:1469 / 1479
页数:11
相关论文
共 21 条
[1]   Contextually Learnt Detection of Unusual Motion-Based Behaviour in Crowded Public Spaces [J].
Arandjelovic, Ognjen .
COMPUTER AND INFORMATION SCIENCES II, 2012, :403-410
[2]  
Arandjelovic O, 2014, LECT NOTES COMPUT SC, V8835, P327, DOI 10.1007/978-3-319-12640-1_40
[3]  
Colmenarez A., 2004, Patent, Patent No. [EP 1 459 272 A1, 1459272A1]
[4]   An improved data stream summary: the count-min sketch and its applications [J].
Cormode, G ;
Muthukrishnan, S .
JOURNAL OF ALGORITHMS-COGNITION INFORMATICS AND LOGIC, 2005, 55 (01) :58-75
[5]  
Cormode G., 2004, P 2004 ACM SIGMOD IN, P35
[6]   Detection of Dynamic Background Due to Swaying Movements From Motion Features [J].
Duc-Son Pham ;
Arandjelovic, Ognjen ;
Venkatesh, Svetha .
IEEE TRANSACTIONS ON IMAGE PROCESSING, 2015, 24 (01) :332-344
[7]   STREAM ORDER AND ORDER STATISTICS: QUANTILE ESTIMATION IN RANDOM-ORDER STREAMS [J].
Guha, Sudipto ;
Mcgregor, Andrew .
SIAM JOURNAL ON COMPUTING, 2009, 38 (05) :2044-2059
[8]  
iCetana, 2014, IMOTIONFOCUS
[9]  
Intellvisions, 2014, IQ PRIS
[10]   THE P2 ALGORITHM FOR DYNAMIC CALCULATION OF QUANTILES AND HISTOGRAMS WITHOUT STORING OBSERVATIONS [J].
JAIN, R ;
CHLAMTAC, I .
COMMUNICATIONS OF THE ACM, 1985, 28 (10) :1076-1085