Theses and Dissertations at Montana State University (MSU)
Permanent URI for this collectionhttps://scholarworks.montana.edu/handle/1/733
Browse
11 results
Filters
Settings
Search Results
Item Calibration and characterization of a VNIR hyperspectral imager for produce monitoring(Montana State University - Bozeman, College of Engineering, 2020) Logan, Riley Donovan; Chairperson, Graduate Committee: Joseph A. Shaw; Joseph A. Shaw was a co-author of the article, 'Measuring the polarization response of a VNIR hyperspectral imager' in the journal 'SPIE proceedings' which is contained within this thesis.; Bryan Scherrer, Jacob Senecal, Neil S. Walton, Amy Peerlinck, John W. Sheppard, and Joseph A. Shaw were co-authors of the article, 'Hyperspectral imaging and machine learning for monitoring produce ripeness' in the journal 'SPIE proceedings' which is contained within this thesis.Hyperspectral imaging is a powerful remote sensing tool capable of capturing rich spectral and spatial information. Although the origins of hyperspectral imaging are in terrestrial remote sensing, new applications are emerging rapidly. Owing to its non-destructive nature, hyperspectral imaging has become a useful tool for monitoring produce ripeness. This paper describes the process of characterizing and calibrating a visible near-infrared (VNIR) hyperspectral imager for obtaining accurate images of produce to be used in machine learning algorithms for analysis. In this work, many calibrations and characterization are outlined, including: a radiance calibration, the process of calculating reflectance, pixel uniformity and image stability testing, spectral characterization, illumination source analysis, and measurement of the polarization response. The images obtained by the calibrated hyperspectral imager were converted to reflectance across a spectral range of 387.12 nm to 1023.5 nm, with a spectral resolution of 2.12 nm. A convolutional neural network was used to perform age classification for Yukon Gold potatoes, bananas, and green peppers. Additionally, a genetic algorithm was used to determine the wavelengths carrying the most useful information for age classification. Experiments were run using red green blue (RGB) images, full-spectrum hyperspectral images, and the wavelengths selected by the genetic algorithm feature selection method. Preliminary data from these analyses show promising results at accurately classifying produce age. The genetic algorithm feature selection method is being used to develop a low-cost multispectral imager for use in monitoring produce in grocery stores.Item Utilizing distributions of variable influence for feature selection in hyperspectral images(Montana State University - Bozeman, College of Engineering, 2019) Walton, Neil Stewart; Chairperson, Graduate Committee: John SheppardOptical sensing has been applied as an important tool in many different domains. Specifically, hyperspectral imaging has enjoyed success in a variety of tasks ranging from plant species classification to ripeness evaluation in produce. Although effective, hyperspectral imaging can be prohibitively expensive to deploy at scale. In the first half of this thesis, we develop a method to assist in designing a low-cost multispectral imager for produce monitoring by using a genetic algorithm (GA) that simultaneously selects a subset of informative wavelengths and identifies effective filter bandwidths for such an imager. Instead of selecting the single fittest member of the final population as our solution, we fit a univariate Gaussian mixture model to a histogram of the overall GA population, selecting the wavelengths associated with the peaks of the distributions as our solution. By evaluating the entire population, rather than a single solution, we are also able to specify filter bandwidths by calculating the standard deviations of the Gaussian distributions and computing the full-width at half-maximum values. In our experiments, we find that this novel histogram-based method for feature selection is effective when compared to both the standard GA and partial least squares discriminant analysis. In the second half of this thesis, we investigate how common feature selection frameworks such as feature ranking, forward selection, and backward elimination break down when faced with the multicollinearity present in hyperspectral data. We then propose two novel algorithms, Variable Importance for Distribution-based Feature Selection (VI-DFS) and Layer-wise Relevance Propagation for Distribution-based Feature Selection (LRP-DFS), that make use of variable importance and feature relevance, respectively. Both methods operate by fitting Gaussian mixture models to the plots of their respective scores over the input wavelengths and select the wavelengths associated with the peaks of each Gaussian component. In our experiments, we find that both novel methods outperform variable ranking, forward selection, and backward elimination and are competitive with the genetic algorithm over all datasets considered.Item Convolutional neural networks for multi- and hyper-spectral image classification(Montana State University - Bozeman, College of Engineering, 2019) Senecal, Jacob John; Chairperson, Graduate Committee: John SheppardWhile a great deal of research has been directed towards developing neural network architectures for classifying RGB images, there is a relative dearth of research directed towards developing neural network architectures specifically for multi-spectral and hyper-spectral imagery. The additional spectral information contained in a multi-spectral or hyper-spectral image can be valuable for land management, agriculture and forestry, disaster control, humanitarian relief operations, and environmental monitoring. However, the massive amounts of data generated by a multi-spectral or hyper- spectral instrument make processing this data a challenge. Machine learning and computer vision techniques could automate the analysis process of these rich data sources. With these benefits in mind, we have adapted recent developments in small efficient convolutional neural networks (CNNs), to create a small CNN architecture capable of being trained from scratch to classify 10 band multi-spectral images, using much fewer parameters than popular deep architectures, such as the ResNet or DenseNet architectures. We show that this network provides higher classification accuracy and greater sample efficiency than the same network using RGB images. We also show that it is possible to employ a transfer learning approach and use a network pre-trained on multi-spectral satellite imagery to increase accuracy on a second much smaller multi-spectral dataset, even though the satellite imagery was captured from a much different perspective (high altitude, overhead vs. ground based at close stand-off distance). These results demonstrates that it is possible to train our small network architectures on small multi-spectral datasets and still achieve high classification accuracy. This is significant as labeled hyper-spectral and multi-spectral datasets are generally much smaller than their RGB counterparts. Finally, we approximate a Bayesian version of our CNN architecture using a recent technique known as Monte Carlo dropout. By keeping dropout in place during test time we can perform a Monte Carlo procedure using multiple forward passes of our network to generate a distribution of network outputs which can be used as a measure of uncertainty in the predictions a network is making. Large variance in the network output corresponds to high uncertainty and vice versa. We show that a network that is capable of working with multi-spectral imagery significantly reduces the uncertainty associated with class predictions compared to using RGB images. This analysis reveals that the benefits of an architecture that works effectively with multi-spectral or hyper-spectral imagery extends beyond higher classification accuracy. Multi-spectral and hyper-spectral imagery allows us to be more confident in the predictions that a deep neural network is making.Item Stigmatic spectroscopy of the solar atmosphere in the vacuum-ultraviolet(Montana State University - Bozeman, College of Letters & Science, 2020) Courrier, Hans Thomas; Chairperson, Graduate Committee: Charles C. Kankelborg; Charles C. Kankelborg was a co-author of the article, 'Using local correlation tracking to recover solar spectral information from a slitless spectrograph' in the journal 'Journal of astronomical telescopes and imaging systems, SPIE' which is contained within this dissertation.; Charles C. Kankelborg, Bart De Pontieu and Jean-Pierre Wulser were co-authors of the article, 'An on orbit determination of point spread functions for the interface region imaging spectrograph' in the journal 'Solar physics' which is contained within this dissertation.; Charles C. Kankelborg, Amy R. Winebarger, Ken Kobayashi, Brent Beabout, Dyana Beabout, Ben Carroll, Jonathan W. Cirtain, James A. Duffy, Carlos Gomez, Eric M. Gullikson, Micah Johnson, Jacob D.Parker, Laurel A. Rachmeler, Roy T. Smart, Larry Springer and David L. Windt were co-authors of the article, 'The EUV snapshot imaging spectrograph (ESIS)' which is contained within this dissertation.The solar atmosphere presents a complicated observing target since tremendous variability exists in solar features over a wide range of spatial, spectral, and temporal scales. Stigmatic spectrographs are indispensable tools that provide simultaneous access to spatial context and spectroscopy, enabling the diagnosis of solar events that cannot be accomplished by imaging or spectroscopy alone. In this dissertation I develop and apply a novel technique for on orbit spectrograph calibration, recover co-temporal Doppler shifts of widely spaced solar features, and describe a new design for a slitless solar spectrograph. The Interface Region Imaging Spectrograph, (IRIS) is currently the highest spatial and spectral resolution, space based, solar spectrograph. Ongoing calibration is important to maintaining the quality of IRIS data. Using a Mercury transit against the backdrop of the dynamic solar atmosphere, I characterize the spatial point spread functions of the spectrograph with a unique, iterative, blind, deconvolution algorithm. An associated deconvolution routine improves the ability of IRIS to resolve spatially compact solar features. This technique is made freely available to the community for use with past and future IRIS observations. The Multi-Order Extreme Ultraviolet Spectrograph (MOSES) is a slitless spectrograph that collects co-temporal, but overlapping spatial and spectral images of solar spectral lines. Untangling these images presents an ill-posed inversion problem. I develop a fast, automated method that returns Doppler shifts of compact solar objects over the entire MOSES field of view with a minimum of effort and interpretation bias. The Extreme ultraviolet Snapshot Imaging Spectrograph (ESIS) is a slitless spectrograph that extends the MOSES concept. I describe this new instrument, which is far more complex and distinct as compared to MOSES, and the contributions I made in the form of optical design and optimization. ESIS will improve the quality of spatial and spectral information obtained from compact and extended solar features, and represents the next step in solar slitless spectroscopy. Taken together, these contributions advance the field by supporting existing instrumentation and by developing new instrumentation and techniques for future observations of the solar atmosphere.Item Development of a smart camera system using a system on module FPGA(Montana State University - Bozeman, College of Engineering, 2017) Dack, Connor Aquila; Chairperson, Graduate Committee: Ross K. SniderImaging systems can now produce more data than conventional PCs with frame grabbers can process in real-time. Moving real-time custom computation as close as possible to the image sensor will alleviate the bandwidth bottle-neck of moving data multiple times through buffers in conventional PC systems, which are also computation bottlenecks. An example of a high bandwidth, high computation application is the use of hyperspectral imagers for sorting applications. Hyperspectral imagers capture hundreds of colors ranging from the visible spectrum to the infrared. This masters thesis continues the development of the hyperspectral smart camera by integrating the image sensor with a field programmable gate array (FPGA) and by developing an object tracking algorithm for use during the sorting process, with the goal of creating a single compact embedded solution. An FPGA is a hardware programmable integrated circuit that can be reprogrammed depending on the application. The prototype integration involves the development of a custom printed circuit board to connect the data and control lines between the sensor, the FPGA, and the control code to read data from the sensor. The hyperspectral data is processed on the FPGA and is combined with the object edges to make a decision on the quality of the object. The object edges are determined using a line scan camera, which provides data via the Camera Link interface, and a custom object tracking algorithm. The object tracking algorithm determines the horizontal edges and center of the object while also tracking the vertical edges and center of the object. The object information is then passed to the air jet sorting subsystem which ejects bad objects. The solution is to integrate the hyperspectral image sensor, the two processing algorithms, and Camera Link interface into a single, compact unit by implementing the design on the Intel Arria 10 System on Module with custom printed circuit boards.Item Hyper-spectral microscope: auto-focusing(Montana State University - Bozeman, College of Engineering, 2018) Lozano, Kora Michelle; Chairperson, Graduate Committee: Ross K. SniderThis thesis is part of a larger project to develop a hyper-spectral microscope, to be used to find the optimal growing conditions for human inducible pluripotent stem cells. The hyper-spectral microscope is being developed by the Department of Chemistry and Biochemistry at Montana State University (MSU). Specifically, the hyper-spectral microscope is being developed to aide in live cell imaging, reduce cell stress from laser excitation, increase the number of markers possible at once, and keep costs down compared to non-hyper-spectral set-ups of similar capability. To the knowledge of those involved in this project it is the first of its kind. The scope of this thesis centers on implementing an auto-focusing algorithm for the hyper-spectral imager.Item Characterization of CpI and CpII [FeFe]-hydrogenases reveals properties contributing to catalytic bias(Montana State University - Bozeman, College of Letters & Science, 2016) Artz, Jacob Hansen; Chairperson, Graduate Committee: John W. Peters; Dissertation contains two articles of which Jacob Hansen Artz is not the main author.; David W. Mulder, Michael W. Ratzloff, Saroj Poudel, Axl X. LeVan, S. Garrett Williams, Michael W. W. Adams, Anne K. Jones, Eric S. Boyd, Paul W. King and John W. Peters were co-authors of the article, 'Potentiometric EPR spectral deconvolution of CPI [FeFe]-hydrogenase reveals accessory cluster properties' submitted to the journal 'Journal of the American Chemical Society' which is contained within this thesis.; David W. Mulder, Michael W. Ratzloff, Saroj Poudel, Axl X. LeVan, Michael W. W. Adams, Eric S. Boyd, Paul W. King, and John W. Peters were co-authors of the article, 'EPR and FTIR spectroscopy provides insights into the mechanism of [FeFe]-hydrogenase CPII' submitted to the journal 'Journal of the American Chemical Society' which is contained within this thesis.The need for food, fuel, and pharmaceuticals has been increasing at a growing rate as the world's population increases and lifestyles improve. All of these needs are highly energy dependent, and, to a significant degree, rely on an inefficient use of fossil fuels. In order to break free of this dependence, new understanding is required for how to efficiently generate the products humanity needs. Here, a model system of two closely related [FeFe]-hydrogenases, CpI and CpII, is employed in order to understand how biology is able to efficiently control the formation of reduced products, in order to further delineate the limits of control, and the extent to which biology may be co-opted for technological needs. CpI, one of nature's best catalysts for reducing protons to hydrogen gas, is compared to CpII, which functions catalytically to oxidize hydrogen to protons and electrons. Oxygen sensitivity, midpoint potentials, catalytic mechanisms, and catalytic bias are explored in-depth using electron paramagnetic resonance, Fourier Transform Infrared spectroscopy, and protein film voltammetry. CpI and CpII have been found to function under different metabolic conditions, and key amino acids influencing their distinct behavior have been identified. The conduit arrays of hydrogenases, which direct electrons to or from the active site, have been found to have distinct midpoint potentials in CpII compared to CpI, effectively reversing the favored electron flow through CpII in comparison to CpI. In order to probe the contributions of the protein framework on catalysis, analysis of site-specific amino acid substituted variants have been used to identify several determinants that affect the H-cluster environment, which contributes to the observed differences between CpI and CpII. This has resulted in a deeper understanding of the hydrogenase model system and the ability to directly influence catalytic bias. Thus, the work presented here represents key progress towards developing unidirectional catalysts, and demonstrates the possibility of targeted, rational design and implementation of unidirectional catalysts.Item Explosive events in the quiet Sun: extreme ultraviolet imaging spectroscopy instrumentation and observations(Montana State University - Bozeman, College of Letters & Science, 2017) Rust, Thomas Ludwell; Chairperson, Graduate Committee: Charles C. KankelborgExplosive event is the name given to slit spectrograph observations of high spectroscopic velocities in solar transition region spectral lines. Explosive events show much variety that cannot yet be explained by a single theory. It is commonly believed that explosive events are powered by magnetic reconnection. The evolution of the line core appears to be an important indicator of which particular reconnection process is at work. The Multi-Order Solar Extreme Ultraviolet Spectrograph (MOSES) is a novel slitless spectrograph designed for imaging spectroscopy of solar extreme ultraviolet (EUV) spectral lines. The spectrograph design forgoes a slit and images instead at three spectral orders of a concave grating. The images are formed simultaneously so the resulting spatial and spectral information is co-temporal over the 20'x10' instrument field of view. This is an advantage over slit spectrographs which build a field of view one narrow slit at a time. The cost of co-temporal imaging spectroscopy with the MOSES is increased data complexity relative to slit spectrograph data. The MOSES data must undergo tomographic inversion for recovery of line profiles. I use the unique data from the MOSES to study transition region explosive events in the He II 304 A spectral line. I identify 41 examples of explosive events which include 5 blue shifted jets, 2 red shifted jets, and 10 bi-directional jets. Typical doppler speeds are approximately 100km s-1. I show the early development of one blue jet and one bi-directional jet and find no acceleration phase at the onset of the event. The bi-directional jets are interesting because they are predicted in models of Petschek reconnection in the transition region. I develop an inversion algorithm for the MOSES data and test it on synthetic observations of a bi-directional jet. The inversion is based on a multiplicative algebraic reconstruction technique (MART). The inversion successfully reproduces synthetic line profiles. I then use the inversion to study the time evolution of a bi-directional jet. The inverted line profiles show fast doppler shifted components and no measurable line core emission. The blue and red wings of the jet show increasing spatial separation with time.Item Defect and coherent transient optical spectroscopy of rare earth doped crystals(Montana State University - Bozeman, College of Letters & Science, 1997) Wang, GuangmingItem Seasonal forage dry matter production and quality of 29 dryland grasses in Montana(Montana State University - Bozeman, College of Agriculture, 2001) Blunt, Kurtis Russell; Chairperson, Graduate Committee: Dennis CashProducers must have accurate and reliable measurements of both forage production and quality in their pastures. Previous studies with dryland grasses in Montana have mostly been limited to adaptation or yield performance of a few species at a single location. The first objective of this study was to document yield and forage quality characteristics of adapted dryland grass varieties over a three-year period at three separate locations. Another objective was to accurately predict forage quality constituents of numerous dryland forage grasses using near infrared spectroscopy (NIRS). A third objective was to generate predictive models for crude protein (CP), acid detergent fiber (ADF), neutral detergent fiber (NDF), and in vitro digestible dry matter (IVDDM) using date or growing degree days. And finally, the last objective of this study was to demonstrate how forage quality information generated in this study can be useful for future improvement of animal carrying capacity predictions. Twenty-nine dryland grass varieties were established at three Montana locations. Data were collected over a three-year period. Forage production and quality data were gathered under a wide range of climatic conditions. Interactions among years, varieties, and locations illustrated the variability of the climate in Montana and biological differences of varieties at different locations. This study makes a strong case for the use of NIRS technology in estimating forage quality of dryland grasses in Montana. Compared to traditional wet chemistry procedures, NIRS proved to be much faster and generated accurate results. Predictive models using date and growing degree days generated estimates of forage quality similar to NIRS but standard errors associated with model parameters limited statistical differences among varieties for season-long forage quality. However, it was determined that the rates of forage quality decline among many of the varieties studied were different (P < 0.01). The r^2 values for predicted forage quality ranged from 0.39 (Rosana, ADF) to 0.85 (Schwendimar, ADF) for the AGGD models and all were highly significant. Strong negative correlations between yield and quality were not found in this study (-0.3<0.3). ADF and NDF were highly correlated (r>0.79). It appears that with optimal management, both forage production and quality can be. optimized. Preliminary use of this data suggests that energy becomes limiting first as the growing season progresses, followed by intake and protein. Further studies should be devoted to modeling pasture carrying capacity with forage quality data.