LOUIS University of Alabama in Huntsville
Not a member yet
8547 research outputs found
Sort by
Creating a 3D Optical LUT for Skin Cancer Classification
No full texthttps://louis.uah.edu/rceu-hcr/1519/thumbnail.jp
Chaos-based thumbnail-preserving image encryption
No full textAs the landscape of image data usage continues to change, cloud services need to store data securely, reliably, and at times maintain a reasonable level of human usability. Thumbnail-preserving encryption (TPE) is a niche area of image encryption research that provides a balance between security and privacy. A TPE scheme encrypts an image in such a way that the thumbnail of the encrypted image is the same as or similar to the thumbnail of the original image. This work presented a chaos-based TPE scheme and explored potential pitfalls encountered when using chaos for encryption. The TPE scheme used a chaos-based encryption key as a pseudorandom number generator for pixel sum preservation and shuffling. The efforts of this thesis aim to further the understanding and implementation of chaos-based TPE and make suggestions on future improvements and for applications in industry
Radar resolution improvements using reference target sliding method in range doppler matrix
Full text linkA novel reference target sliding method (RTSM) in the frequency domain of post-match-filtered radar data is presented, and it is shown that target resolution can be improved to levels much finer than classical bandwidth-limited resolution metrics. The RTSM uses a priori knowledge of the radar system to test returned signals against baseline references to provide expected target fitting. The methods developed have broad application and can be used to provide crossing target resolution and estimation parameter refinement in the presence of overlapping targets. Additionally, the reduced requirement for bandwidth in target resolution will allow for low-bandwidth systems to perform on par with higher-bandwidth systems in several applications. The proposed methods are also well-suited to improve detection and tracking performance in lower cost arrays such as non-actively scanned arrays that employ digital beamforming. The RTSM is presented from an underlying mathematical approach, where input reference data can be formed from closed-form equations or with discrete methods. Tracking case studies and statistical analysis show that RTSM improves target crossing performance in a large variety of cases. Material is presented that demonstrates the algorithm and methods are robust enough to not require strict point targets as input. While RTSM is currently set to accurately resolve no more than two overlapped targets, test data are presented that show its ability to continue to work with relatively high accuracy with interferers present. Finally, discussion is offered for the future direction of this work, including the expansion to a first-look Doppler ambiguity resolver
Correlating optical phase and atmospheric coherence width in characterizing turbulence
Full text linkThis dissertation presents the development, validation, and optimization of a novel Hartmann Turbulence Sensor (HTS) for real-time, high-speed characterization of atmospheric optical turbulence. The HTS captures and processes live data at 1 kHz and reports turbulence parameters at 22 Hz. A key innovation is the integration of a dominant spectral component selection method into the Difference in Differential Tilt Variance (DDTV) algorithm. By using spectral analysis to identify areas of interest (AOIs) based on their dominant spatial frequency content, this method reduces computational load by up to 75% while maintaining a low RMS error. It also offers a visual representation of clustered dominant spectral components, providing insights into the cumulative turbulent structures along the propagation path. Experimental validation demonstrated that the HTS accurately measures essential turbulence parameters, and that the integration of dynamic AOI selection with dominant spectral analysis provides a novel approach to real-time adaptive sensing of optical turbulence
F-region neutral wind responses to magnetospheric forcing in the nightside auroral oval
Full text linkHigh-latitude neutral winds have a number of drivers, both from solar and magnetospheric origins as well as from the lower atmosphere. The efficiency of the individual thermospheric forces, in addition to the relative strength of ion-neutral coupling, are not well characterized. Because of this, the neutral wind response to changes in ionospheric convection is not well understood, with response times ranging on the order of tens of minutes to hours. Using both data driven and modeling simulation efforts, we aim to develop a better understanding of high-latitude ionosphere-thermosphere coupling. In order to do this, three individual studies will be done to answer the following questions: 1) What is the time-scale of neutral wind\u27s response to changes in magnetospheric and auroral forcing? 2) How does this ion-neutral coupling efficiency change with varying levels of geomagnetic activity? 3) What is the relative contribution of each of the thermospheric forces, and what forces dominate neutral wind responses? Using data from Scanning Doppler Imagers (SDIs), the Poker Flat Incoherent Scatter Radar (PFISR), and All Sky Imagers (ASIs), a new neutral wind response time calculation method is developed. This method differs from previous calculation methods, such as the e-folding time and time-lagged correlation analysis, by providing a time evolution of the response time that considers all thermospheric forces. Using this new response time calculation, we perform a statistical analysis of neutral wind responses to magnetospheric forcing. Then, we investigate how this response time changes with respect to geomagnetic activity level by performing a superposed epoch analysis on the short and long response events to uncover what geomagnetic conditions lead to faster or longer response times. This analysis will be done using the AE index, SYM/H index, OMNI IMF data, local magnetometer data from the THEMIS ground array and electron density from PFISR. For our modeling efforts, the thermosphere-ionosphere-electrodynamics global circulation model (TIE-GCM) is used to investigate the influence of various high-latitude ionospheric drivers of thermospheric winds during a storm period. Individual thermospheric forcing terms are also calculated in order to identify dominant drivers of the neutral winds
Why believe an unreliable jailhouse informant? : the role of prosecutorial vouching in juror decision-making
Full text linkThe present study asked jurors to explain why they believe jailhouse informant (JI) testimony, and investigated the role of prosecutorial vouching (i.e., the assumption that any witness called by the prosecution is credible) in jurors’ decision-making. A total of 89 juryeligible participants read a trial summary in which the JI was either absent, reliable (i.e., testified to consistent and accurate crime facts), or unreliable (i.e., accurate facts during trial, but inaccurate facts in initial statement). Primary results were that (a) in open-ended responses, participants assumed that the prosecution would not allow a dishonest or uncredible JI to testify (i.e., prosecutorial vouching), (b) the unreliable JI reduced conviction rates (vs the reliable JI), but these rates were still higher than the no-JI condition, and (c) in open-ended responses, participants across reliability conditions believed the JI because of details in the JI’s testimony, despite acknowledging the unreliable JI’s inconsistencies
Evaluating bias in facial recognition datasets : a study on representation and classification fairness
Full text linkFacial recognition technology is widely used in security and identity verification but exhibits biases that disproportionately impact certain demographic groups. This thesis examines racial and skin tone biases in commonly used facial recognition datasets, assessing their influence on classification models. Using manual classification via the Fitzpatrick scale and numerical skin tone representations from LAB color values, we identified significant dataset imbalances, with lighter skin tones (Fitzpatrick Types 1 and 2) overrepresented and darker tones (Types 4-6) underrepresented. Model evaluation using Random Forest, SVM, and XGBoost showed an overall accuracy of 72\%, but classification for Asian individuals was notably weaker (0.45 precision, 0.36 recall), suggesting dataset bias affects performance. Comparisons with US Census 2020 data further revealed discrepancies in racial representation, underscoring the need for better dataset alignment. To mitigate these issues, this study highlights the importance of manual auditing and human-in-the-loop methodologies in ensuring fairness. Future research should focus on curating balanced datasets, employing fairness-aware algorithms, and analyzing bias propagation within facial recognition models to advance equitable AI development
