One of the fundamental problems in computer vision is automatically tracking an object’s trajectory over the time. Formally called “Generic Visual Object Tracking”, this field has accumulated decades of research and significant progress has recently been made. However, majority of the literature has focused on tracking objects on regular video streams with Red-Green-Blue data channels. While methods that are proposed for this type of data have achieved great success, they are limited by their nature since they do not have access to the depth data of the scene. This problem causes the loss of vital information on the object’s true 3D structure and makes it more difficult to detect certain, common phenomenon in tracking such as occlusion.
On the other hand, the last decade has seen the introduction and adoption of cheap 3D sensors into many aspects of our lives, including our mobile phones. Thanks to this, we can now easily obtain high quality RGBD data in which a fourth depth channel encodes the real physical distances of the objects in the scene. Hence, it is now possible to model a greater understanding of the environment using a video stream.
In his thesis, Uğur Kart has proposed novel tracking algorithms and a dataset which can track an object in RGBD video while being aware of occlusions and disappearances.
“These have applications in various industries such as sports, military, surveillance and robotics,” Uğur Kart says.
Uğur Kart is currently working as a Computer Vision Engineer at Scandit Tampere where he develops computer vision algorithms for enterprises.
The doctoral dissertation of M.Sc. (Tech) Uğur Kart in the field of Signal Processing titled Visual Object Tracking on RGBD Videos Using Discriminative Correlation Filters will be publicly examined in the Faculty of Information Technology and Communication Sciences on Friday 8th of October 2021 in the auditorium TB109 of Tietotalo at Tampere university at 12:00 PM (Korkeakoulunkatu 1, Tampere). The Opponent will be Assoc. Prof. Ville Kyrki from Aalto University and the Custos will be Prof. Joni Kämäräinen from Faculty of Information Technology and Communication Sciences.
Limited number of participants are invited due to COVID-19 restrictions. The dissertation can be followed remotely through Panopto remote connection.
The dissertation is available online at http://urn.fi/URN:ISBN:978-952-03-2058-4.
The faculty administration provides a pdf-version of the thesis at request: itc.tau [at] tuni.fi
Photo: Aslı Pekman