Object tracking in augmented reality remote access laboratories without fiducial markers

Smith, Mark (2019) Object tracking in augmented reality remote access laboratories without fiducial markers. [Thesis (PhD/Research)]

[img]
Preview
Text (Whole Thesis)
M Smith 23 Jul 19.pdf

Download (8MB) | Preview

Abstract

Remote Access Laboratories provide students with access to learning resources without the need to be in-situ (with the assets). The technology endows users with access to physical experiments anywhere and anytime, while also minimising or distributing the cost of operation for expensive laboratory equipment. Augmented Reality is a technology which provides interactive sensory feedback to users. The user experiences reality through a computer-based user interface with additional computer-generated information in the form applicable to the targeted senses.

Recent advances in high definition video capture devices, video screens and mobile computers have driven resurgence in mainstream Augmented Reality technologies. Lower cost and greater processing power of microprocessors and memory place the resources in the hands of developers and users alike, allowing education institutes to invest in technologies that enhance the delivery of course content. This increase in pedagogical resources has already allowed the phenomenon of education at a distance to reach students from a wide range of demographics, improving access and outcomes in multiple disciplines. Incorporating Augmented Reality into Remote Access Laboratories resources has the benefit of improving overall user immersion into the remote experiment, thus improving student engagement and understanding of the delivered material.

Visual implementations of Augmented Reality rely on providing the user with seamless integration of the current environment (through mobile device, desktop PC, or heads up display) with computer generated artificial visual artefacts. Virtual objects must appear in context to the current environment, and respond in a realistic period, or else the user suffers from a disjointed and confusing blend of real and virtual information. Understanding and interacting with the visual scene is controlled through Computer Vision algorithms, and are crucial in ensuring that the AR systems co-operate with the data discovered through the systems.

While Augmented Reality has begun to expand in the educational environment, currently, there is still very little overlap of Augmented Reality technologies with Remote Access Laboratories. This research has investigated Computer Vision models that support Augmented Reality technologies such that live video streams from Remote Laboratories are enhanced by synthetic overlays pertinent to the experiments. Orientation of synthetic visual overlays requires knowledge of key reference points, often performed by fiducial markers. Removing the equipment’s need for fiducial markers and a priori knowledge simplifies and accelerates the uptake and expansion of the technology.

These works uncover hybrid Computer Vision models which require no prior knowledge of the laboratory environment, including no fiducial markers or tags to track important objects and references. Developed models derive all relevant data from the live video stream and require no previous knowledge regarding the configuration of the physical scene. The new image analysis paradigms, (Two-Dimensional Colour Histograms and Neighbourhood Gradient Signature) improve the current state of markerless tracking through the unique attributes discovered within the sequential video frames. Novel methods are also established, with which to assess and measure the performance of Computer Vision models. Objective ground truth images minimise the level of subjective interference in measuring the efficacy of CV edge and corner detectors.

Additionally, locating an effective method to contrast detected attributes associated with an image or object, has provided a means to measure the likelihood of an image match between video frames. In combination with existing material and new contributions, this research demonstrates effective object detection and tracking for Augmented Reality systems within a Remote Access Laboratory environment, with no requirement for fiducial markers, or prior knowledge of the environment. The models that have been proposed in the work can be generalised to be used in any cyber-physical environment that facilitates peripherals such as cameras and other sensors.


Statistics for USQ ePrint 39498
Statistics for this ePrint Item
Item Type: Thesis (PhD/Research)
Item Status: Live Archive
Additional Information: Doctor of Philosophy (PhD) thesis.
Faculty/School / Institute/Centre: Current - Faculty of Health, Engineering and Sciences - School of Mechanical and Electrical Engineering (1 Jul 2013 -)
Faculty/School / Institute/Centre: Current - Faculty of Health, Engineering and Sciences - School of Mechanical and Electrical Engineering (1 Jul 2013 -)
Supervisors: Kist, Alexander; Maxwell, Andrew
Date Deposited: 04 Sep 2020 02:18
Last Modified: 21 Apr 2021 02:15
Uncontrolled Keywords: augmented reality, remote access laboratories, computer vision, image Processing, Object Tracking, Human Computer Interface
Fields of Research (2008): 08 Information and Computing Sciences > 0801 Artificial Intelligence and Image Processing > 080199 Artificial Intelligence and Image Processing not elsewhere classified
Identification Number or DOI: doi:10.26192/XSFE-4A84
URI: http://eprints.usq.edu.au/id/eprint/39498

Actions (login required)

View Item Archive Repository Staff Only