R-3 Repository :: Browsing by Author "Kelly, Kevin"

Browsing by Author "Kelly, Kevin"

Now showing 1 - 12 of 12

A Compressive Sensing and Unmixing Scheme for Hyperspectral Data Processing
(2011-01) Li, Chengbo; Sun, Ting; Kelly, Kevin; Zhang, Yin
Hyperspectral data processing typically demands enormous computational resources in terms of storage, computation and I/O throughputs, especially when real-time processing is desired. In this paper, we investigate a low-complexity scheme for hyperspectral data compression and reconstruction. In this scheme, compressed hyperspectral data are acquired directly by a device similar to the single-pixel camera based on the principle of compressive sensing. To decode the compressed data, we propose a numerical procedure to directly compute the unmixed abundance fractions of given endmembers, completely bypassing high-complexity tasks involving the hyperspectral data cube itself. The reconstruction model is to minimize the total variational of the abundance fractions subject to a pre-processed fidelity equation with a significantly reduced size, and other side constraints. An augmented Lagrangian type algorithm is developed to solve this model. We conduct extensive numerical experiments to demonstrate the feasibility and efficiency of the proposed approach, using both synthetic data and hardware-measured data. Experimental and computational evidence obtained from this study indicates that the proposed scheme has a high potential in real-world applications.
A Spectrum-based Regularization Approach to Linear Inverse Problems: Models, Learned Parameters and Algorithms
(2015-04-20) Castanon, Jorge Castanon Alberto; Zhang, Yin; Tapia, Richard; Hand, Paul; Kelly, Kevin
In this thesis, we study the problem of recovering signals, in particular images, that approximately satisfy severely ill-conditioned or underdetermined linear systems. For example, such a linear system may represent a set of under-sampled and noisy linear measurements. It is well-known that the quality of the recovery critically depends on the choice of an appropriate regularization model that incorporates prior information about the target solution. Two of the most successful regularization models are the Tikhonov and Total Variation (TV) models, each of which is used in a wide range of applications. We design and investigate a class of spectrum-based models that generalize and improve upon both the Tikhonov and the TV methods, as well as their combinations or so-called hybrids. The proposed models contain "spectrum parameters" that are learned from training data sets through solving optimization problems. This parameter-learning feature gives these models the flexibility to adapt to desired target solutions. We devise efficient algorithms for all the proposed models and conduct comprehensive numerical experiments to evaluate their performance as compared to established models. Numerical results show a generally superior quality in recovered images by our approach from under-sampled linear measurements. Using the proposed algorithms, one can often obtain much improved quality at a moderate increase in computational time.
Chemical tuning of electrical and magnetic properties in the transition metal dichalcogenides
(2019-04-19) Choe, Jesse; Morosan, Emilia; Kelly, Kevin; Natelson, Douglas
Transition metal dichalcogenides are a diverse class of layered materials. Due to their quasi two-dimensional nature, they are a sandbox for investigating low dimensional physics, but can be doped in a variety of ways. Not only can substitutional doping occur on either the transition metal or chalcogen site, but intercalation between the layers can tune the system as well. Here I report on the results of three transition metal dichalcogenide systems with three drastically different results. In the Copper-Platinum-Selenium system, initial results suggest two new superconductors in the ternary phase diagram. Doping platinum into TiSe$_2$ results in an increase in the resistivity of several order of magnitude. Angle-resolved photo-emission spectroscopy shows that Platinum doping induces a pseudo gap in the system. Scanning tunneling microscopy measurements show domain wall formation. Power law fits to the resistivity suggests that the electrical transport is dominated by Luttinger liquid behavior. Finally, the intercalation of Iron into TiS$_2$ gives rise to several magnetic features. Large bowtie magnetoresistance arises showing an increase of up to 40\%. Hysteresis in magnetization shows sharp switching behavior coinciding with the bowtie in magnetoresistance. Ferromagnetic order is seen in conjunction with glassy behavior. These results are compared and contrasted to the results in Fe$_x$TiS$_2$. My results in these systems show that the depth and breadth of physical phenomena in the transition metal dichalcogenide system make it a fascinating system for investigating strongly correlated systems.
Exploiting compressive matrices for dynamic infrared object tracking
(2016-04-22) Chen, Jianbo; Kelly, Kevin
Recent development on compressive sensing (CS) presents a great potential for this technique to be used in broader applications from hyper-spectroscopy microscopy to homeland security. And the new mathematics of CS has drastically benefited this field especially in imaging and video applications. Based on novel theoretical principles and experiments, it has been demonstrated that an image can be reconstruct with only K << N measurements from an N-dimensional basis, which is much less than the sampling rate required by the Shannon-Nyquist sampling theorem. The compressive single pixel camera is one embodiment of such an imaging system and has proven capable of capturing both static images and dynamic scenes using fewer measurements than the current schemes. In this thesis we will explore compressive dynamic scene acquisition with prior information or models, incorporating with different sensing matrixes. We demonstrate through simulations and experiments the effectiveness of knowledge-enhanced patterns over unbiased compressive measurements in a variety of applications including motion tracking and object recognition. We also present using a SPC like system for high-speed anomaly detection. Despite its importance in a wide variety of machine vision applications, extending anomaly detection and tracking beyond the visible spectrum in a cost-effective manner presents a significant technological challenge. As a step in this direction, we present a compressive imaging system, specially designed patterns, and a set of metrics to identify the existence of short durance anomalies against a complex background. Our novel measurement design is chosen to be most sensitive to singular anomalies based on the Walsh-Hadamard transform. We illustrate the utility of our approach via a series of simulations and experiments on the compressive single-pixel camera system.
High resolution light field capture using GMM prior and sparse coding
(2014-10-07) Tambe, Salil; Veeraraghavan, Ashok; Sabharwal, Ashutosh; Kelly, Kevin
Light fields, being inherently a 4D function cannot be mapped onto the 2D sensor in a single image without loosing out on resolution. A natural way to overcome this barrier is to capture multiple images to record the light field. However, this method only works for static scenes, therefore the resolution problem stays unresolved, it only gets transformed from the domain of low spatio-angular resolution to a problem of low temporal resolution. In this work, we leverage the redundant nature of light fields to recover them at higher resolution by first capturing a set of well-chosen images, and later reconstructing the LF from these images using some prior-based algorithms. We achieve this in two ways. In the first method, we capture multiplexed light field frames using an electronically tunable programmable aperture and later recover the light field using a motion-aware dictionary learning and sparsity based reconstruction algorithm. The number of adjacent multiplexed frames to be used during the recovery of each light field frame is decided based on the applicability of the static scene assumption. This is determined using optical-flow and forms the basis of our motion-aware reconstruction algorithm. We also show how to optimize the programmable aperture patterns using the learned dictionary. Our second method utilizes focus stacks to computationally recover light fields post-capture [1] . However our method differs from [1] in the following ways. (i) We obtain the entire focus-aperture (45 focus and 18 aperture settings) stack by capturing just a few (about $8-16$) images and computationally reconstructing images corresponding to all other focus-aperture settings, while [1] capture the entire focus stack corresponding to a given aperture setting (ii) Since we recover the focus stack at smaller aperture settings as well, we can produce LFs at finer angular resolutions. We call our method 'Compressive Epsilon Photography' since we capture few (compressive) images with slightly varying parameters (Epsilon Photography) and post-capture computationally reconstruct images corresponding to all other missing parameter combinations. The recovered LF has spatial resolution corresponding to the sensor resolution of the camera and can recover any angular view which lies inside the aperture. [1] A. Levin and F. Durand, "Linear view synthesis using a dimensionality gap light field prior," in IEEE conf. Computer Vision and Pattern Recognition, pp. 1831-1838, 2010
Integrating Coherent Anti-Stokes Raman Scattering Imaging and Deep Learning Analytics for High Precision, Real Time, Label Free Cancer Diagnosis
(2017-08-11) Weng, Sheng; Kelly, Kevin; Wong, Stephen T.C.
Coherent anti-Stokes Raman scattering (CARS) imaging technique has demonstrated great potential in clinical diagnosis by providing cellular-level resolution images without using exogenous contrast agents. This thesis contributes to the formation of an optical fiber based signal collection scheme and an automated image analytics platform to translate CARS microscopy for clinical uses. First, I introduce the concept of CARS by showing original images acquired from thyroid and parathyroid tissues. Second, I describe the use of a customized optical fiber bundle to collect and differentiate forward and backward generated CARS signals that contain different structural information. Third, I demonstrate the feasibility of using deep learning algorithms to characterize and classify CARS images automatically. In particular, I apply transfer learning on the CARS images and achieve 89.2% prediction accuracy in differentiating normal, small-cell carcinoma, adenocarcinoma, and squamous cell carcinoma human lung images. The combination of an optical fiber based microendoscopy and deep learning image classification algorithm will facilitate CARS imaging for on-the-spot cancer diagnosis, allowing medical practitioners to obtain essential information in real time and accelerate clinical decision-making. Meanwhile, the thesis also shows the generality of the deep learning algorithm developed by classifying screening images generated in drug discovery. As an example, for automated classification of large volumes of high-content screening images for Alzheimer’s disease drug discovery, by applying similar transfer learning method on hyperphosphorylated tau images, I categorize drug hits into ineffective, partially-effective, and significantly-effective groups with high speed and accuracy.
Label-free Imaging of Thyroid and Parathyroid Glands Using Coherent Anti-Stokes Raman Scattering (CARS) Microscopy
(2015-04-28) Weng, Sheng; Kelly, Kevin; Wong, Stephen; Thomann, Isabell; Kono, Junichiro
Thyroid and parathyroid glands play a vital role in regulating the body's metabolism and calcium levels. Surgical removal of the glands is the main treatment for both thyroid cancer and parathyroid adenoma. In thyroidectomy and parathyroidectomy, it's very important to differentiate thyroid, parathyroid, and the other tissues around the neck. Traditionally, physicians use ultrasound guided fine needle aspiration (FNA) to evaluate thyroid nodules, but up to 30% of FNA results are “inconclusive”. The sestamibi scan can localize parathyroid adenoma, but currently it only has 50% accuracy. Here we applied the emerging CARS technique to image both thyroid and parathyroid tissues, which has potential to be used in real-time in vivo examination of different structures. We also developed algorithms to differentiate different cellular structures based on CARS images. When incorporated with a fiber optic endoscope in the future, CARS imaging technique can help surgeons identify cancerous thyroid tissue intraoperatively, preserve good parathyroid glands during thyroidectomy and find parathyroid adenoma during parathyroidectomy.
Layered transition metal pnictides investigated by experimental and computational methods
(2015-04-23) Wang, Jiakui; Morosan, Emilia; Dai, Pengcheng; Kelly, Kevin
Intensive research interest on transition metal pnictide compounds was stimulated by the discovery of unconventional superconductivity in Fe-pnictide compounds in 2008. A key observation in Fe-pnictide compunds is the intimate relationship between the structure, magnetism and superconductivity in those compunds. It is thus important to investigate a couple of transition metal pnictide compounds with similar structures to superconducting Fe-pnictide compounds, to elucidate how the magnetic and superconducting properties evolve with the type of crystallographic structure, lattice paramters and elements on a speci c site. In this thesis, R3T4As4O2 (R = La - Sm, T = Ni - Cu), which has a convolutional structure of 122 and 1111 Fe-pnictide superconductors, was investigated from the perspective of a couple of physical properties investigation as well as theoritical computation. The combination of experimental and theoretical methods reveals complex magnetic properties as the elements on rare earth and transition metal sites varies. Such observations would bene t a full understanding of the relationship between structure, magnetism and superconductivity in transition metal pnictide system as well the search for conventional superconductors with this speci c structure.
Magnetization and magnetoresistance in iron intercalated transition metal dichalcogenides
(2016-04-26) Choe, Jesse; Morosan, Emilia; Kelly, Kevin
The understanding of magnetism in strongly correlated electronic systems is a vital area of research. Not only is it linked to other phenomena like high temperature superconductivity in the cuprates and iron pnictides, but magnetic materials have been used in electronics since before the computer. As it becomes harder to prop up Moore's law by increasing the density of transistors, mankind must look towards new methods to improve technology or risk stagnation. Research into alternative materials for technology, such as transition metal dichalcogenides, is a promising direction of research to maintain the rate of technological improvement. Our work focuses on the effect of iron intercalation in TiS$_2$. Single crystals of Fe$_x$TiS$_2$ (0 $\le x \le$ 1) were grown using vapor transport. Anisotropic susceptibility and magnetization measurements of the samples were measured, showing ferromagnetism and sharp switching behavior in the magnetization. Finally electrical transport measurements were taken, both with and without field. Measurements of magnetoresistance for $x$ = 0.2 and 0.3 show large magnetoresistance (up to $\sim$ 60\%) and an atypical `bowtie' shape.
RF Shot Noise Measurements in Au Atomic-scale Junctions
(2016-01-29) Chen, Ruoyu; Natelson, Douglas; Du, Ruirui; Kelly, Kevin
Conduction electrons are responsible for many physical or chemical phenomena in condensed matter systems, and their behavior can be directly studied by electronic transport measurements. In conventional transport measurements, conductance or resistance is usually the focus. Such a measurement can be as simple as a quick two terminal DC check by a multi-meter, or a more sophisticated lock-in measurement of multiple higher harmonic signals synchronized to different frequencies. Conductance carries direct information about the quasi-particle density of states and the local electronic distributions, which are usually Fermi-Dirac distribution. Conductance is modified or dominated by scattering from defacts or interfaces, and could also reflect the spin-spin exchange interactions or inelastic couplings with phonons and photons. Naturally one can ask the question: is there anything else we can measure electronically, which carries extra information that a conductance measurement does not provide? One answer to this question is the electronic noise. While the conductance reflects the average charge conduction ability of a system, noise describes how the physical quantities fluctuate around their average values. Some of the fluctuations carry information about their physical origins. This thesis will focus on one particular type of the electronic noise shot noise, but other types of noise will also be introduced and discussed. We choose to measure the radio frequency component of shot noise, combining with a modulated lock-in detection technique, which provides a method to largely get rid of other unwanted low-frequency noise signals. Au atomic-scale junctions are the systems we studied here. Au is relatively well understood and will not generate too many complications, so it's ideal as the first platform for us to understand both shot noise itself and our RF technique. On the other hand, the atomic scale raises fundamental questions about electronic transport and local energy exchange and dissipation, which make our measurements fundamentally interesting. We employed two different types of mechanical controlled Au break junctions: the Scanning Tunneling Microscope(STM)-style Au break junctions, and the mechanically bending Au break junctions. We studied shot noise behaviors of individual configurations or ensemble averages over all the accessible configurations. Measurements were conducted at both room temperature and liquid He temperature. High quality shot noise measurements were demonstrated. New phenomena like anomalous excess noise enhancement at high bias voltages and non-zero shot noise variance below 1G0 were seen. We also found shot noise to be surprisingly insensitive to temperatures between 4.2K and 100K, and can be well described by the non-interacting approximation.
Terahertz Spectroscopy of Soft Condensed Matter: Single Methyl Branched n-Alkanes and Plastic-Crystals
(2015-02-25) Nickel, Daniel Vincent; Mittleman, Daniel M.; Kelly, Kevin; Natelson, Douglas
Terahertz time-domain spectroscopy has long been recognized as a valuable tool for characterizing the far-IR range (0.1 THz to 10 THz) properties of innumerable condensed phase materials. Soft condensed matter, a broad subcategory of condensed matter, encompasses thermally and/or mechanically deformable phases such as liquids, liquid-crystals, plastic-crystals, and polymers. With comparatively more internal degrees of freedom and weaker atomic/molecular interactions, complex phase behavior and dynamics often emerge in these materials, which generally have not been explored in the THz range. Here, temperature-dependent THz-TDS is applied to three soft condensed matter systems. Using a custom designed reflection geometry spectrometer and a novel variable path length liquid sample cell, the isotropic polarizabilities of liquid phase methyl-branched n-alkanes and their linear structural isomers are characterized and compared in the THz range. In addition, the plastic-crystal’s succinonitrile and camphor are studied using transmission geometry THz-TDS. The THz range properties and dynamics characterized in these studies, from the structure-dependent isotropic polarizabilities of prototypical liquid hydrocarbons to the complex disorder/order transitions and correlated molecular dynamics of plastic-crystals, provide valuable fundamental insight into the behavior of soft condensed matter systems.
Wafer-scale films of aligned single-wall carbon nanotubes: preparation, characterization, and optoelectronic applications
(2015-11-20) He, Xiaowei; Kono, Junichiro; Kelly, Kevin; Hauge, Robert; Adams, Wade
Single-wall carbon nanotubes (SWCNTs) are one-dimensional materials defined by a cylindrical and hollow structure with aspect ratios of up to 10^7:1. Individual SWCNTs have been shown to possess excellent electric, optical, thermal, and mechanical properties that are promising for electronic and optoelectronic device applications. However, when they are assembled into macroscopic objects such as films and fibers, these unique properties tend to vanish, primarily due to disorder. Hence, methods are being sought for fabricating ordered SWCNT assemblies for the development of high-performance devices based on SWCNTs. In this dissertation, we present two methods for preparing highly aligned SWCNT films with excellent optoelectronic properties. The first method is based on vertically aligned SWCNT arrays grown by water-assisted chemical vapor deposition. We transferred these arrays to desired substrates to form horizontally aligned SWCNT films and created p-n junction devices that worked as flexible, room-temperature-operating, and polarization-sensitive infrared and terahertz photodetectors. The second method is based on our discovery of spontaneous global alignment of SWCNTs that occurs during vacuum filtration of SWCNT suspensions. By carefully controlling critical factors during vacuum filtration, we obtained wafer-scale, monodomain films of strongly aligned SWCNTs. By measuring polarization-dependent terahertz transmittance, we demonstrated ideal polarizer performance with large extinction ratios. The universality of this method was confirmed by applying it to diverse types of SWCNTs, all of which showed exceptionally high degrees of alignment. Furthermore, we successfully fabricated aligned SWCNT films enriched in one specific chirality by combining our new method with an advanced nanotube sorting technique: aqueous two-phase extraction. Transistors fabricated using such films showed very high conductivity anisotropies and excellent on-off ratios.

Browsing by Author "Kelly, Kevin"

Results Per Page

Sort Options