학술논문

Object tracking using fuzzy-based improved graph, interesting patches and multi-label MRF optimization.
Document Type
Article
Source
Multimedia Systems. Jun2023, Vol. 29 Issue 3, p1431-1451. 21p.
Subject
*IMAGE representation
*OBJECT tracking (Computer vision)
*FUZZY logic
*FUZZY neural networks
*NOISE
Language
ISSN
0942-4962
Abstract
The complete object-tracking system typically comprises four phases. In this paper, we aim to focus on the first three phases. Graphs are often used in image representation by organizing feature points. In this study, there are three areas of discussion. One is that the number of points is large and using all the points complicates the problem. The number of selected points should be limited. The second is which nodes are related to each other. The third issue is that graphs created from images are usually not stable because of noise or deformation. Therefore, the inaccurate matching of the graph must be obtained. We have also focused on these issues in this paper and propose to unify these steps. An experimental study is performed over several sequences with many challenging situations. A comparison with other methods reports that the proposed approach is more advantageous. [ABSTRACT FROM AUTHOR]