Journal of Applied Sciences ›› 2015, Vol. 33 ›› Issue (5): 502-517.doi: 10.3969/j.issn.0255-8297.2015.05.005

• Signal and Information Processing • Previous Articles     Next Articles

Multi-feature Selection Tracking Based on Support Vector Machine

HU Zhao-hua1,2, XU Yu-wei1,3, ZHAO Xiao-lei1, HE Jun1, ZHOU You4   

  1. 1. School of Electronic & Information Engineering, Nanjing University of Information Science and Technology, Nanjing 210044, China;
    2. Jiangsu Collaborative Innovation Center on Atmospheric Environment and Equipment Technology, Nanjing 210044, China;
    3. 94804 PLA Troops, Shanghai 200434, China;
    4. Suzhou Electrical Apparatus Science Research Institute Co., Ltd, Suzhou 215000, Jiangsu Province, China
  • Received:2015-04-21 Revised:2015-05-26 Online:2015-09-30 Published:2015-09-30

Abstract: Discriminative tracking is generally based on a single feature and uses the current track result (a positive sample) and some negative samples to train the classifier. It may lead to tracking due to occlusion, illumination changes and deformation. To overcome the problem caused by single feature description and single positive sample training, we propose a multi-feature selection algorithm based on the support vector machine (SVM). The classifier is trained with multiple positive and negative samples based on multi-feature descriptions. In the tracking step, a candidate sample with maximum confidence probability is used as the tracking result. Further, we take advantage of a subspace learning method to update positive samples used in the classifier training. Experimental results show that the proposed algorithm has desirable performance in a variety of challenging situations.

Key words: classifier, object tracking, support vector machine (SVM), subspace learning, multi-feature selection

CLC Number: