published_or_final_version / abstract / toc / Computer Science and Information Systems / Master / Master of Philosophy

published_or_final_version / Industrial and Manufacturing Systems Engineering / Doctoral / Doctor of Philosophy

Purpose: To determine whether semi-scleral contact lenses (sSCL) can be appropriately fitted using corneal sagittal depth measurements, and to determine the impact of fit on visual acuity, effect of time on topographic corneal clearance and comfort ratings. Method: Three sSCL (Jupiter 15mm; Essilor) were fit to 20 subjects who had previous diagnoses of KC (n=18) or PMD (n=2). The fitting of the sSCL were based on the CSD measured with the Visante™ OCT at a 15mm chord on the horizontal meridian. To select the sSCL from the diagnostic trial lens set, values of 325 (lens 1), 375 (lens 2) and 425 (lens 3) μm were randomly added in sequence to the CSD. Subjects were allowed to wear each of the sSCL for 1hour. After this time, the central corneal clearance (CCC) was assessed using an UL-OCT, high contrast visual acuity (HCVA) and low contrast visual acuity (LCVA) were measured using a LogMAR VA chart and comfort ratings were obtained using a comfort rating scale (0-100). Results: The mean CSD in the horizontal meridian was 3.78±0.53 (range: 3.33-4.17) mm at a 15mm chord. The mean CCC was 190±100, 360±120 and 450±170 µm for each lens respectively (p=0.001). The mean CCC loss was 30.00±40.00, 30±60.00 and 40.00±50 µm for each lens respectively (p>0.05). The mean HCVA for lenses 1, 2 and 3 were 0.05±0.12, 0.07±0.11 and 0.11±0.08 respectively, which were significantly different (p=0.02). Tukey post hoc analysis demonstrated that this difference was only significant between lenses 1 and 3 (p=0.01). Similar findings were found for LCVA. The overall comfort rating for all three sSCL was 77.7±10.6. The comfort ratings for lenses 1, 2 and 3 were 74.9±9.2, 79.7 ±11.6 and 78.6±10.8 respectively. These differences were not significantly different (p=0.24). Conclusion: Evaluation of CSD can be used effectively to select which sSCL to fit on the eye. The results of this study suggest that lens 2 (adding 375 μm to the CSD) gave the best combination of VA and comfort ratings. However, evaluation of the fluorescein pattern must be balanced with the VA and comfort ratings for successful fitting of sSCL in a clinical setting. There was also a likelihood of topographic corneal loss after 1 hour of sSCL wear; however, this may vary depending on many factors such as scleral zone and its relationship with the scleral conjunctiva. Eyelid force, design of the contact lens and other unknown factors may play a part in the contact lens settling time and amount.

Two recent studies suggest that Inhibition of Return (IOR) varies in magnitude as a function of target location for overt orienting tasks but not covert orienting tasks. Unfortunately, methodological differences between these studies prevent a direct comparison of their results. Thus the aim of the current study was to replicate and extend the results of these two studies within a single experiment while controlling for methodological differences. Participants (N=37) were assigned to a cue-target or a target-target group and were required to make manual (covert orienting block) or saccadic responses (overt orienting block) to peripheral stimuli occupying one of four peripheral locations. An analysis of target reaction times indicated that while IOR was present under all circumstances, it did not vary as a function of target location. A careful examination of our methods points to the importance of controlling set size (the number of possible target locations) in IOR studies.

At the heart of every stereo vision algorithm is a solution to the matching problem - the problem of finding points in the right and left image that correspond to a single point in the real world. Applying assumptions regarding the epipolar rectification and color similarity between two frames is often not possible for real-world image capture systems, like those used in urban search and rescue robots. More flexible and robust feature descriptors are necessary to operate under harsh real world conditions. This thesis compares the accuracy of disparity images generated using local features including points, line segments, and regions, as well as a global framework implemented using loopy belief propagation. This thesis will introduce two new algorithms for stereo matching using line segments and regions, as well as several support structures that optimize the algorithms performance and accuracy. Since few complete frameworks exist for line segment and region features, new algorithms that were developed during the research for this thesis will be outlined and evaluated. The comparison includes quantitative evaluation using the Middlebury stereo image pairs and qualitative evaluation using images from a less structured environment. Since this evaluation is grounded in urban search and rescue robotics, processing time is a significant constraint which will be evaluated for each algorithm. This thesis will show that line segment-based stereo vision with a gradient descriptor achieves at least a 10% better accuracy than all other methods used in this evaluation while maintaining the low runtime associated with local feature based stereo vision.

At the heart of every stereo vision algorithm is a solution to the matching problem - the problem of finding points in the right and left image that correspond to a single point in the real world. Applying assumptions regarding the epipolar rectification and color similarity between two frames is often not possible for real-world image capture systems, like those used in urban search and rescue robots. More flexible and robust feature descriptors are necessary to operate under harsh real world conditions. This thesis compares the accuracy of disparity images generated using local features including points, line segments, and regions, as well as a global framework implemented using loopy belief propagation. This thesis will introduce two new algorithms for stereo matching using line segments and regions, as well as several support structures that optimize the algorithms performance and accuracy. Since few complete frameworks exist for line segment and region features, new algorithms that were developed during the research for this thesis will be outlined and evaluated. The comparison includes quantitative evaluation using the Middlebury stereo image pairs and qualitative evaluation using images from a less structured environment. Since this evaluation is grounded in urban search and rescue robotics, processing time is a significant constraint which will be evaluated for each algorithm. This thesis will show that line segment-based stereo vision with a gradient descriptor achieves at least a 10% better accuracy than all other methods used in this evaluation while maintaining the low runtime associated with local feature based stereo vision.

Individuals with autism spectrum disorder (ASD) often have difficulty “seeing the big picture” and tend to process objects by their parts. This study used a perceptual grouping task (Experiment 1) involving the Ponzo illusion to determine if individuals with ASD show a local processing bias due to difficulties grouping stimuli preattentively. Individuals with ASD were less likely than TD individuals to report an illusion-based response. The percentage of responses consistent with the illusion indicated at chance performance in the ASD group, suggesting that they experience deficits preattentively when grouping stimuli. This study also used a viewing window paradigm (Experiment 2) to evaluate the parts-based processing strategies used by individuals with ASD when allowed to either actively or passively view blurry objects using a restricted viewing aperture. For both conditions performance was similar across groups, suggesting that individuals with ASD use similar parts-based processing strategies as TD individuals to identify objects.

Automated visual recognition and detection processes are becoming increasingly prevalent in almost all scientific fields and being currently implemented in many fields of industry. In most cases, systems are painstakingly designed and developed in order to detect only a single and specific object or property of an object. The objective of this project was to create a framework of development in which any object distinguishable in a two-dimensional digital image could be analyzed and subsequently detected in other images. Furthermore, as new methods are developed, they could be easily incorporated into this framework to ultimately improve the performance of the system. / This thesis describes a highly adaptable, general-application visual detection system as well as several innovative methods for the description of objects without which such adaptivity would be impossible. Two-dimensional, still images are analyzed and objects of interest can be introduced to the system. Objects are then described by a variety of properties through derived attributes and stored in a database. Occurrences of these objects can then be detected in future images through comparisons to selected models. The system is fully expandable in that new properties and comparison techniques or criteria can be added as they are developed and as their need becomes apparent. The system is presented with a basic set of attribute representations and methods of comparison, and their development and origin are described in detail. The database structure is outlined and the process by which new properties and comparative methods can be added is described. Seventeen different images containing nearly two thousand separate objects were searched for various model objects and the average classification accuracy was 98.3%. In most images, more than 100 object classifications could be performed per second at an accuracy higher than 95% when no higher order analyses were required.

This thesis explores three object tracking algorithms for image sequences. These algorithms include the ensemble tracker, the EM-like mean-shift colour-histogram tracker, and the wandering-stable-lost scale-invariant feature transform (WSL-SIFT) tracker. The algorithms are radically different from one another. Despite their differences, they are evaluated on the same publicly available, moderately sized, research data sets which include 129 test cases in 13 different scenes. The results aid in fostering an understanding of their respective behaviours and in highlighting their flaws and failures. Lastly, an implementation setup is described that is suited to large-scale, grid computing, batch testing of these algorithms. Results clearly indicate that none of the evaluated trackers are suited to general purpose use. However, one may intelligently choose a tracker for a well-defined application by analysing the known scene characteristics.

Typescript. / Thesis (Ph. D.)--University of Hawaii, 1966. / Bibliography: leaves 81-83. / ix, 83 l illus., tables