[1]杨彪,邹凌,李文杰,等.基于自适应加权二部图的多特征目标匹配[J].常州大学学报(自然科学版),2015,(03):66-69.[doi:10.3969/j.issn.2095-0411.2015.03.013]
　YANG Biao,ZOU Ling,LI Wenjie,et al.Multi-Features Object Matching Based on Adaptive Weighted Bipartite Graph[J].Journal of Changzhou University(Natural Science Edition),2015,(03):66-69.[doi:10.3969/j.issn.2095-0411.2015.03.013]
点击复制

基于自适应加权二部图的多特征目标匹配()
分享到：

常州大学学报(自然科学版)[ISSN:2095-0411/CN:32-1822/N]

卷:
期数:
2015年03期
页码:
66-69
栏目:
出版日期:
2015-07-25

文章信息/Info

Title:
Multi-Features Object Matching Based on Adaptive Weighted Bipartite Graph
作者:
杨彪1; 2; 邹凌1; 2; 李文杰1; 2; 周天彤1; 2
1.常州大学 信息科学与工程学院，江苏 常州 213164；2.常州市生物医学信息技术重点实验室，江苏 常州 213164
Author(s):
YANG Biao1; 2; ZOU Ling1; 2; LI Wenjie1; 2; ZHOU Tiantong1; 2
1.School of Information Science and Engineering, Changzhou University, Changzhou 213164,China; 2. Changzhou Key Laboratory of Biomedical Information Technology, Changzhou 213164, China
关键词:
多特征; 目标匹配; 自适应加权二部图; Kuhn-Munkres算法; MAP问题
Keywords:
multiple features; object matching; adaptive weighted bipartite graph; Kuhn-Munkres algorithm;  MAP problem
分类号:
U 491
DOI:
10.3969/j.issn.2095-0411.2015.03.013
文献标志码:
A
摘要:
目标匹配是在大范围多摄像机监控网络中进行连续目标跟踪的基础，对无重叠视野多摄像机网络中的目标匹配进行研究，提出了一种基于自适应加权二部图的多特征目标匹配算法。考虑到不同摄像机视野下成像角度、光照的差异，采用多特征融合技术构造目标的观测模型，并利用贝叶斯准则将目标匹配问题表示成最大后验概率（MAP）问题。同时，提出一种自适应加权二部图替代MAP问题，并利用Kuhn-Munkres算法解出二部图的最大权匹配。通过对监控数据进行试验，表明本文算法可在接受的时间范围内改善目标匹配的准确度。
Abstract:
Object matching is the basis for continuous object tracking under wide area monitoring using camera network. This paper focuses on object matching across non-overlapping camera views. A multi-features object matching approach based on adaptive weighted bipartite graph is proposed. Multiple features are employed to construct an observation model due to view variation and illumination change across different camera views. Object matching is then represented as a maximum a posteriori (MAP) problem based the Bayesian rule. Meanwhile, the MAP problem is replaced using an adaptive weighted bipartite graph which is then solved by the Kuhn-Munkres algorithm. Experimental results under realistic camera network indicate that our approach can improve the accuracy of object matching across non-overlapping camera views within an acceptable time.

参考文献/References:

[1]CAI Y H, HUANG K Q, TANG T N. Human appearance matching across multiple non-overlapping cameras [C]. Tampa: IEEE Computer Society, 2008.
[2]ZHAO R, OUYANG W L, WANG X G. Unsupervised Salience Learning for Person Re-identication [C]. Portland: IEEE Computer Society, 2013.
[3]JAVED O, SHAFIQUE K. Modeling inter-camera space-time and appearance relationships for tracking across non-overlapping views [J]. Computer Vision and Image Understanding. 2008, 109(2):146-162.
[4]刘少华,赖世铭,张茂军. 基于最小费用流模型的无重叠视域多摄像机目标关联算法 [J]. 自动化学报, 2010, 36(10):1484-1489
[5]LI S H, GUO G Z. The application of improved HSV color space model in image processing [C]. Wuhan: IEEE Computer Society, 2010.
[6]PICCARDI M, CHENG E D. Track matching over disjoint camera views based on an incremental major color spectrum histogram [C]. Como: IEEE Computer Society, 2005.
[7]陈炳权, 刘宏立.基于颜色纹理和形状特征相结合的图像检索系统 [J].湖南文理学院学报(自然科学版), 2009, 21(4):67-70.
[8] HUANG J G, KONG B. A New Method of Unstructured Road Detection Based on HSV Color Space and Road Features [C]. Jeju City: IEEE Computer Society, 2007.
[9]贾科勤，王洪元，程起才.基于自适应阈值Canny算子的视频文本定位方法[J].常州大学学报（自然科学版），2014，26（1）：47-51.
[10] 刘少华,张茂军,陈旺. 无重叠视域多摄像机的数据关联算法 [J]. 计算机应用, 2009, 29(9): 2378-2382.

备注/Memo

备注/Memo:
国家自然科学基金项目（61201096）；常州市科技项目（CE20145055）；江苏省青蓝工程资助
更新日期/Last Update: 2015-11-26