Back to search

Rochester Institute of Technology

Accelerated Object Tracking with Local Binary Features

Abstract

dc:description.abstract

Multi-object tracking is a problem with wide application in modern computing. Object tracking is leveraged in areas such as human computer interaction, autonomous vehicle navigation, panorama generation, as well as countless other robotic applications. Several trackers have demonstrated favorable results for tracking of single objects. However, modern object trackers must make significant tradeoffs in order to accommodate multiple objects while maintaining real-time performance. These tradeoffs include sacrifices in robustness and accuracy that adversely affect the results. This thesis details the design and multiple implementations of an object tracker that is focused on computational efficiency. The computational efficiency of the tracker is achieved through use of local binary descriptors in a template matching approach. Candidate templates are matched to a dictionary composed of both static and dynamic templates to allow for variation in the appearance of the object while minimizing the potential for drift in the tracker. Locality constraints have been used to reduce tracking jitter. Due to the significant promise for parallelization, the tracking algorithm was implemented on the Graphics Processing Unit (GPU) using the CUDA API. The tracker's efficiency also led to its implantation on a mobile platform as one of the mobile trackers that can accurately track at faster than realtime speed. Benchmarks were performed to compare the proposed tracker to state of the art trackers on a wide range of standard test videos. The tracker implemented in this work has demonstrated a higher degree of accuracy while operating several orders of magnitude faster.

Degree

thesis:*
Name thesis:degree_name
Computer Engineering (MS)
Level thesis:degree_level
Thesis
Discipline thesis:degree_discipline
Computer Engineering (KGCOE)

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Minnehan, Breton Lawrence
Contributors dc:contributor
  • Andreas Savakis

Subjects

dc:subject × 3

Identifiers

dc:identifier.*
Repository record dc:identifier
https://repository.rit.edu/theses/8343

Chain of custody

source
Harvested from
Rochester Institute of Technology
Base URL
repository.rit.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Minnehan, Breton Lawrence. Accelerated Object Tracking with Local Binary Features. Thesis thesis, https://repository.rit.edu/theses/8343