Abstract

With the growing ubiquity of mobile devices, users are turning to their smartphones and tablets to perform more complex tasks than ever before. Performing computer vision tasks on mobile devices must be done despite the constraints on CPU performance, memory, and power consumption. One such task for mobile devices involves object tracking, an important area of computer vision. The computational complexity of tracking algorithms makes them ideal candidates for optimization on mobile platforms.

This thesis presents a mobile implementation for real time object tracking. Currently few tracking approaches take into consideration the resource constraints on mobile devices. Optimizing performance for mobile devices can result in better and more efficient tracking approaches for mobile applications such as augmented reality. These performance benefits aim to increase the frame rate at which an object is tracked and reduce power consumption during tracking.

For this thesis, we utilize binary descriptors, such as Binary Robust Independent Elementary Features (BRIEF), Oriented FAST and Rotated BRIEF (ORB), Binary Robust Invariant Scalable Keypoints (BRISK), and Fast Retina Keypoint (FREAK). The tracking performance of these descriptors is benchmarked on mobile devices. We consider an object tracking approach based on a dictionary of templates that involves generating keypoints of a detected object and candidate regions in subsequent frames. Descriptor matching, between candidate regions in a new frame and a dictionary of templates, identifies the location of the tracked object. These comparisons are often computationally intensive and require a great deal of memory and processing time.

Google's Android operating system is used to implement the tracking application on a Samsung Galaxy series phone and tablet. Control of the Android camera is largely done through OpenCV's Android SDK. Power consumption is measured using the PowerTutor Android application. Other performance characteristics, such as processing time, are gathered using the Dalvik Debug Monitor Server (DDMS) tool included in the Android SDK. These metrics are used to evaluate the tracker's performance on mobile devices.

Library of Congress Subject Headings

Computer vision; Automatic tracking--Data processing; Optical pattern recognition

Publication Date

5-2014

Document Type

Thesis

Student Type

Graduate

Degree Name

Computer Engineering (MS)

Department, Program, or Center

Computer Engineering (KGCOE)

Advisor

Andreas Savakis

Advisor/Committee Member

Andres Kwasinski

Advisor/Committee Member

Roy Melton

Comments

A physical copy is available from RIT's Wallace Library at TA1634 .Q87 2014

Campus

RIT – Main Campus

Plan Codes

CMPE-MS

Download

Share

COinS