Multi-sphere classification algorithm based on 1-SVM

被引：0
|
作者：
Wang Hongbo [1 ]
Zhao Guangzhou [1 ]
Qu Yilong [2 ]
Qi Donglian [1 ]
机构：
[1] Zhejiang Univ, Coll Elect Engn, Hangzhou 310027, Zhejiang, Peoples R China
[2] Minist Sci & Technol, High Technol Res & Dev Ctr, Beijing 100044, Peoples R China
来源：
PROCEEDINGS OF THE 31ST CHINESE CONTROL CONFERENCE | 2012年
关键词：
1-SVM; multi-sphere; multiple classifiers; license plate recognition;
D O I：
暂无
中图分类号：
TP [自动化技术、计算机技术];
学科分类号：
0812 ;
摘要：
The multi-sphere 1-SVM classifier(MSVDD) is proposed by expending the 1-SVM to multi-class model. When classifying, in order to ensure that boundaries are obtained using 1-SVM tight enough for characterizing all classes in a given data set, introduce implicitly decomposition tactics, and propose MSVDD method in two different mathematical formulations in this paper. The simulation results show that the algorithm can be efficiently applied to License Plate Recognition System.
引用
收藏
页码：3989 / 3995
页数：7
相关论文
共 12 条
  • [1] [Anonymous], TECHNICAL REPORT
  • [2] [Anonymous], 1998, MULTI CLASS SUPPORT
  • [3] A comparison of methods for multiclass support vector machines
    Hsu, CW
    Lin, CJ
    [J]. IEEE TRANSACTIONS ON NEURAL NETWORKS, 2002, 13 (02): : 415 - 425
  • [4] Kressel UHG, 1999, ADVANCES IN KERNEL METHODS, P255
  • [5] Platt JC, 2000, ADV NEUR IN, V12, P547
  • [6] Schölkopf B, 2000, ADV NEUR IN, V12, P582
  • [7] Support vector domain description
    Tax, DMJ
    Duin, RPW
    [J]. PATTERN RECOGNITION LETTERS, 1999, 20 (11-13) : 1191 - 1199
  • [8] Support vector data description
    Tax, DMJ
    Duin, RPW
    [J]. MACHINE LEARNING, 2004, 54 (01) : 45 - 66
  • [9] Vapnik V., 1998, Statistical Learning Theory, P5
  • [10] Structured one-class classification
    Wang, Defeng
    Yeung, Daniel S.
    Tsang, Eric C. C.
    [J]. IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART B-CYBERNETICS, 2006, 36 (06): : 1283 - 1295
12