Show/Hide Menu
Hide/Show Apps
anonymousUser
Logout
Türkçe
Türkçe
Search
Search
Login
Login
OpenMETU
OpenMETU
About
About
Açık Bilim Politikası
Açık Bilim Politikası
Frequently Asked Questions
Frequently Asked Questions
Browse
Browse
By Issue Date
By Issue Date
Authors
Authors
Titles
Titles
Subjects
Subjects
Communities & Collections
Communities & Collections
Conjoint individual and group tracking framework with online learning
Download
index.pdf
Date
2016
Author
Yiğit, Ahmet
Metadata
Show full item record
Item Usage Stats
1
views
0
downloads
A group is a social unit which consists of people interacting with each other and sharing the similar characteristics. Because of social properties of group, group tracking re-quires taking into account not only visual properties but also social properties such as interaction of people with each other. Also, people groups are dynamic entities and they may grow and shrink with merge and split events. This dynamic nature makes it difficult to track groups using conventional trackers. Besides these difficulties, differ-ent types of groups require different strategies in order to perform tracking effectively. While it is possible to track individuals separately when group is sparse, group is con-sidered as a single entity when it is dense. To overcome and address these challenges, we propose a new tracking strategy, named the Conjoint Individual and Group Tracking (CIGT), based on particle filter and online learning from discriminative appearance model in this thesis. The CIGT proposes a multi observation model with in-group and out-group weights in order to track groups and to evaluate merge and split events. CIGT has two complementary phases: tracking and learning. In the tracking phase, the CIGT calculates multiple weights from observa-tions and models individuals and groups with merge and split events. Particle advection is used in the motion model of CIGT to facilitate tracking of dense groups. In the learn-ing phase, reliable tracklets are first created. Then discriminative appearance model, consisting of shape, color and texture features, is extracted and used in AdaBoost online learning. State estimate is performed for both individuals and groups by using the discriminative learning model.
Subject Keywords
Machine learning.
,
Artificial intelligence.
,
Computer algorithms.
,
Social interaction.
,
Social groups.
URI
http://etd.lib.metu.edu.tr/upload/12619809/index.pdf
https://hdl.handle.net/11511/25517
Collections
Graduate School of Informatics, Thesis