Search

Video tracking involves estimating the state(s) of target(s) over time on the image plane, where the sequence of target states is termed as a trajectory. Trajectory evaluation refers to the assessment of a tracker's results that may be based on quantification of the discrepancy in the estimated states with respect to the corresponding ground-truth states. In this thesis, after presenting a review of the related work, we make the following proposals for the ground-truthbased trajectory evaluation in videos. We propose three overlap-based measures that account for the key aspects of multi-target tracking evaluation including accuracy, cardinality error and ID changes. The measures quantify tracking performance by combining accuracy and cardinality errors at frame level, computing the sequence-level tracking accuracy at varying accuracy levels, and measuring ID changes while considering the length of the track in which they occur. An extensive experimental validation of the proposed measures is conducted using four state-of-the-art multi-target trackers on challenging real-world publicly-available datasets. The proposed measures show advantages (because they are parameter independent and numerically bounded) over the existing measures and enable a thorough evaluation of trackers while identifying their strengths and weaknesses. We present a protocol that is composed of a set of trials that evaluate the robustness of trackers on a range of test scenarios representing several real-world conditions. To compare single-target trackers' performance on trials, we present a single-score parameter-independent evaluation measure that quantifies tracking success and failure, and combines them for both summative and formative performance assessment. The protocol is validated on publicly available sequences with a diversity of targets and challenges using eight state-of-the-art single-target trackers. Through an extensive experimental analysis, the framework facilitates the selection of trackers for different operational conditions in real-world applications and for different target types. Finally, to quantitatively compare the relative performances of evaluation measures we propose a methodology based on determining the probabilistic agreement between tracking result decisions made by measures and those made by humans. We use tracking results on publicly available datasets with different target types and varying challenges, and collect the judgments of 90 skilled, semi-skilled and unskilled human subjects using a web-based performance assessment test. The analysis of the agreements allows us to highlight the variation in performance of the different measures and the most appropriate ones for the evaluation and comparison of trackers.

As deep space missions expand in scope and distance, the efficiency of propulsion systems becomes paramount. This thesis analyzes the impact of small measurement errors in the thrust profiles of Hall-Effect Thrusters, a common type of electric propulsor, known for increased efficiency compared to traditional chemical propulsion despite lower thrust. Due to their prolonged operational times, these errors compound, affecting trajectory and mission success. Through analyzing the AEPS Hall-Effect Thruster prototype, designed for NASA’s Artemis Program’s Gateway space station, using curve fitting and a Monte Carlo simulation, we assess the effects of these errors on an example mission to Alpha Centauri. Results show plasma dynamics cause the majority of error, but cause minimal trajectory deviation and propellant loss. This reinforces electric propulsion’s suitability for long-distance space travel. This work informs spacecraft mission design, providing valuable insights into fuel efficiency and system selection, as well as building upon NASA Glenn Research Center’s Electric Propulsion and Power Laboratory’s prior research.

This work discusses the development of and subsequent use of a novel method for the measurement of stability derivatives of models in a hypersonic gun tunnel using the free-flight technique. The technique uses a passive electro-optical system to illuminate and scan a rectangular field of view in the vertical plane through the test-section of the gun tunnel. 6 vertical laser sheets, produced by the expansion and reflection of 3 He-Ne laser sources, are used to backlight the model as it flies through the field of view. Shadows from sheets interrupted by the model are projected onto rapid-access photo-sensitive arrays. The position and pitch angle of a model traversing the field of view can be determined by a knowledge of the recorded shadow positions and the external geometry. The non-linear differential equations of motion are fitted to the resulting motions using a parametric differentiation technique embedded in a modified least-squares minimisation scheme. Convergence of the minimisation procedure yields the stability terms which are of interest. The effects of model geometry and mass distribution on the resulting free-flight static and dynamic stability are assessed and discussed.

This thesis aims to investigate the potential of a transfer orbit using successive aerocaptures at Saturn and Titan to establish a science orbit around Titan. Titan is an Earth-like moon with a dense atmosphere and organic compounds present. It has many similarities with Earth that are useful to study such as superrotation. Superrotation is when the atmosphere rotates faster than the body it surrounds. In order to study Titan, we need to establish an orbit around it. The Saturn system is distant from Earth, 8.5 Astronomical Units (AU) which makes it difficult to reach from a time and velocity point of view. We propose to use an aerocapture at Saturn to intercept Titan with lower relative velocity in order to perform an aerocapture at Titan. The analysis was performed in primarily MATLAB to simulate the orbits. The results of this showed that we can aerocapture a spacecraft at Saturn and arrive at Titan within roughly 4 to 8 km/s relative velocity regardless of the incoming hyperbolic excess velocity at the Saturn system. This can be improve upon by using intermediate transfer orbits, such as bi-elliptics, to arrive with even lower relative velocities to Titan of as low as 1 km/s. The drag acceleration experienced during the Saturn aerocapture had peak values of between 0.2 and 1.4 g's and acceleration over 50% of the peak is experienced between 6.8 and 8 minutes. This capture method has the potential to make Titan more easily accessible and allow for scientific study of a clear target for improving our understanding of Earth-like processes on other bodies in our solar system.

Trajectory Prediction is the problem of predicting the short-term and long-term spatialcoordinates of various agents such as cars, buses, pedestrians. In Trajectory Prediction we use themoment of an agent’s past trajectories to predict action or course the agent is going to make in thefuture. It is difficult to predict trajectory of a player since various factors after the decision whicha player makes, leading to his next step. We propose a model which uses video input from asport, then extracting the data from the video to extract the players movement pattern from whichthe trajectory of the player is predicted through trajectory prediction Network. We implementobject detection and efficient tracking to extract the players position information from the dataset.Our model has a Mean Average Displacement (MAD) of 8.35

The relationship between parental education and family income to parenting practices and children’s achievement is well-documented. The effects of high parental education on these behaviors in the face of financial difficulty over time, however, are less established. Therefore, this study seeks to examine the longitudinal income status trajectories of low-income, highly-educated families in addition to the predictors and outcomes of experiencing varying durations of economic hardship. The first research aim identified income status trajectories of college-educated families living in or near poverty. The second research aim determined whether sociodemographic characteristics, such as race or sex, predict to different durations of economic hardship. The third research aim examined the effect of different durations of economic hardship on distal parent and child outcomes. The last research aim attempted to replicate findings in a conceptually similar dataset. Two datasets were used to investigate these research aims: Early Child Longitudinal Study 1998-99 (ECLS-K: 1998) and the Early Childhood Longitudinal Study 2010-11 (ECLS-K: 2010). To be included in the analyses, families must be low-income (i.e., living at or below 200% of the federal poverty line) and have at least one parent with a college degree (i.e., Bachelor’s degree or higher) when their child was in kindergarten. Families must also have participated in at least two waves of data following kindergarten (ECLS-K: 1998 N = 540; ECLS-K: 2010: N = 449). First, latent growth curve analyses (LCGA) in Mplus identified income status trajectories. Second, multinomial logistic regressions determined predictors of the previously identified latent trajectories. Last, mean equality tests examined distal outcome differences amongst the previously identified latent trajectories. These three analyses were first conducted in the ECLS-K: 1998 and then replicated in the ECLS-K: 2010. In the ECLS-K: 1998, two trajectories emerged: Transient (i.e., above the low-income threshold in most waves) and Chronic (i.e., at or below the low-income threshold in most waves). Compared to the Transient class families, the Chronic class families were more likely to have a parent identify as Black and live in rural areas and less likely to have a parent with an occupation requiring a postsecondary degree. Parents in the Transient class reported higher school involvement, satisfaction with their child’s school, and warmth towards their child and children in the Transient class had higher reading and math achievement compared to families in the Chronic class. In the ECLS-K: 2010, three trajectories emerged: Immediate Transient (i.e., immediately above the low income threshold in most waves), Delayed Transient (i.e., initially at or below the low-income threshold but gradually rose above it over time), and Chronic (i.e., at or below the low-income threshold in most waves). No sociodemographic characteristics predicted trajectory class membership. Parents in both the Delayed Transient and Chronic classes reported more rules in the household than parents in the Immediate Transient class. This study was one of the first to investigate income trajectories in low-income, highly-educated families over time. Variation in duration of poverty sociodemographic characteristics and outcomes amongst low-income, college-educated families highlight the need for a range of policy and program solutions to address diverse familial needs. Parental education may be a protective buffer for parents and children in the face of economic hardship. The difference in results between the two datasets suggests issues of historical validity and additional sensitivity analyses identified important methodological differences when comparing across datasets.

While traditional accounts characterize speech planning as serial, discrete, and walled off from behavior, newer proposals describe it as parallel, continuous, and delicately interweaved with execution. This dissertation evaluates the claim that speakers under time pressure produce articulatory trajectories that are graded fits to continually evolving partial, parallel constraints. In three form preparation experiments, motion-tracked digital video captured oral postures over time. Participants were sometimes primed with the nuclear vowels of CVC words. All words contained either /i/ (spread lips; e.g., “seed”) or /u/ (rounded lips, e.g., “soup”). In the first two experiments, initial consonants were highly uncertain. In Experiment 1, articulatory postures produced before stimulus presentation anticipated the upcoming vowel when it was primed. Experiment 2 included two pairs of vowel-primed blocks. The words of one /i/-block began with the rounded consonants /w/ or /r/ (e.g., “weak”). Primed vowels influenced anticipatory articulation except when superseded by conflicting demands from likely initial consonants. In Experiment 3, initial consonants were always bilabial, but words were blocked such that either vowel (/i/ versus /u/) or coda type (alveolar/lateral versus bilabial/labiodental) was initially uncertain. For example, a participant might know that an upcoming word would use an /i/ sound but not know whether the final consonant left the mouth relatively open (e.g., “meal”) or closed (e.g., “peep”). Analysis emphasized words ending in alveolar/lateral codas, and followed the evolution of articulatory trajectories over the acoustic phase of the utterance, in both horizontal and vertical lip aperture dimensions. Contrary to expectations, priming did not appreciably impact the shape of these two-dimensional trajectories.

Flocculation is critical in drinking water treatment; in flocculation, the particle size distribution changes from a large number of small particles to a small number of larger particles. Larger particles are effectively removed by settling and filtration processes that follow flocculation. In recent years, manufacturing of engineered nanoparticles has skyrocketed, and these nanoparticles can enter our water supplies, but knowledge of their fate in water treatment is limited. The objective of this research was to update knowledge of flocculation by extending previous work at the microscale to the nanoscale. Flocculation involves transport of particles to the vicinity of one another and subsequent attachment if interactions at close distances are favorable. Transport and possible collisions are brought about by Brownian motion, differential sedimentation, and fluid shear; these processes, even at the nanoscale, are well understood. Whether collisions and attachment actually occur, however, depend on a balance of hydrodynamic interactions, van der Waals attraction, and electrostatic repulsion; this research quantitatively assessed, for the first time, this balance for collisions of nanoparticles by all three collision mechanisms using a well-established trajectory analysis approach. The collision efficiency (α) is the ratio of the number of successful collisions (attachment) to the number of collisions predicted by the transport equations. The analysis was performed with and without electrostatic repulsion, which occurs if particles are charged. In all cases, Brownian motion was the dominant flocculation mechanism. However, without electrostatic repulsion, differential sedimentation and fluid shear were found to be far more important than heretofore expected because the α value can be substantially higher than one, contrary to all previous understanding. With electrostatic repulsion, collisions by these two mechanisms were found to occur only if the particles are substantially different in size. Experiments in which the changes in particle size distributions of nanoparticles were carefully monitored were also performed, and the results compared to the mathematical predictions. Although not perfect, excellent agreement between the measured and predicted particle size distributions was found. The conclusion is optimistic: if nanoparticles are properly destabilized to reduce or eliminate surface charge, they will be well removed in conventional water treatment plants.

This thesis is set in the context of solving the problem of user’s cognitive workload in real-time, which is a complex multifaceted problem. Attempts have been made to measure cognitive workload with single and multimodal input with some success - but no comprehensive framework has existed earlier in this regard. The investigation begins by comprehensively reviewing efforts to measure cognitive workload in both human factors and psychology of learning domains. Human factors domain is relevant to the real-time dimension and application of this problem while psychology of learning provides the comprehensive cognitive load theory framework for experimental research. With this background, the thesis explores user mouse input modality and develops temporal and spatial features that can index user’s cognitive workload under experimental conditions. These features are then further explored for (a) generalizability using three subject groups and (b) viability in real-time environment using modified adaptive windowing and concept drift detection techniques. A pause/hesitation analysis is developed to link user mouse interactivity features to changing cognitive workload levels. Considering these empirical explorations and other prior attempts, a framework is proposed for real-time cognitive workload measures. User’s Load-Effort Homeostasis (LEH) is an important discovery while putting together the framework and is then explored using Big-Five personality traits. Another contribution is explaining and interpreting LEH using both Baddeley’s working memory model and Cowen’s embedded process model. Framework applicability is then demonstrated by applying it to the problem of making machine learning usable (in the context of decision support systems). Finally, interactive machine learning is presented and argued as the best application context for the real-time cognitive workload framework. Future work would include developing and extending the multimodal aspect of this framework and integrating it more concretely with the interactive machine learning approach.

We have a huge amount of video data from extensively available surveillance cameras and increasingly growing technology to record the motion of a moving object in the form of trajectory data. With proliferation of location-enabled devices and ongoing growth in smartphone penetration as well as advancements in exploiting image processing techniques, tracking moving objects is more flawlessly achievable. In this work, we explore some domain-independent qualitative and quantitative features in raw trajectory (spatio-temporal) data in videos captured by a fixed single wide-angle view camera sensor in outdoor areas. We study the efficacy of those features in classifying four basic high level actions by employing two supervised learning algorithms and show how each of the features affect the learning algorithms’ overall accuracy as a single factor or confounded with others.

Manual video analysis is apparently a tedious task. An efficient solution is of highly importance to automate the process and to assist operators. A major goal of video analysis is understanding and recognizing human activities captured by surveillance cameras, a very challenging problem; the activities can be either individual or interactional among multiple objects. It involves extraction of relevant spatial and temporal information from visual images. Most video analytics systems are constrained by specific environmental situations. Different domains may require different specific knowledge to express characteristics of interesting events. Spatial-temporal trajectories have been utilized to capture motion characteristics of activities. The focus of this dissertation is on how trajectories are utilized in assist in developing video analytic system in the context of surveillance. The research as reported in this dissertation begins real-time highway traffic monitoring and dynamic traffic pattern analysis and in the end generalize the knowledge to event and activity analysis in a broader context. The main contributions are: the use of the graph-theoretic dominant set approach to the classification of traffic trajectories; the ability to first partition the trajectory clusters using entry and exit point awareness to significantly improve the clustering effectiveness and to reduce the computational time and complexity in the on-line processing of new trajectories; A novel tracking method that uses the extended 3-D Hungarian algorithm with a Kalman filter to preserve the smoothness of motion; a novel camera calibration method to determine the second vanishing point with no operator assistance; and a logic reasoning framework together with a new set of context free LLEs which could be utilized across different domains. Additional efforts have been made for three comprehensive surveillance systems together with main contributions mentioned above.

This dissertation describes research investigating the potential for using Landsat data to identify and characterize woody canopy cover on reclaimed coal-mined lands through three separate studies. The objective of the first study was to assess whether surface coal mines in the forested central Appalachian regions of the US can be separated from the other prevalent forest-replacing disturbances through analysis of an interannual chronosequence of Landsat images. Disturbances were classified using descriptors of the disturbance/recovery trajectories: disturbance minimum, recovery slope and recovery maximum. Three vegetation indices (VIs) (normalized difference vegetation index, NDVI; tasseled cap greenness/brightness ratio, TC G/B; and inverse of Landsat band 3, B3I) were used to analyze multitemporal trajectories generated using both pixels and objects. Classification accuracies using objects were better than those obtained using pixels for all VIs. The highest object-based classification accuracy was achieved using TC G/B (89%), followed by NDVI (88%) and B3I (80%). The objective of the second study was to evaluate performance of a woody canopy cover (including both native and invasive species) estimation method based on the 2011 National Land Cover Database (NLCD) protocol for both mined and non-mined areas of the central Appalachians. Potential explanatory variables included raw and derived bands from leaf-on and leaf-off Landsat scenes plus terrain descriptors. Results show that the model developed to estimate canopy cover for mines (R2 = 0.78, Adj. R2 = 0.77, RMSE = 16%) is more robust than the models developed for non-mines, mixed, and all areas combined. The objective of the third study was to determine whether four disturbance/recovery parameters (recovery time, disturbance minimum, recovery slope, and recovery maximum), alone or in combination with variables identified in the second study, enable robust estimation of woody canopy cover on reclaimed surface coal mines. Of the disturbance/recovery parameters, only recovery time made a significant contribution to the model (R2 0.45, Adj. R2 0.44, RMSE 14%). Addition of leaf-on and leaf-off NDVI improved the R2 to 0.54 (Adj. R2 0.53, RMSE 13%). Analysis of Landsat data has strong potential for identifying reclaimed mines and characterizing the extent to which woody canopy has recovered post-reclamation.

Abstract: Compact facility designs and retro-fitting of facilities that involve heavy lifts are often performed in congested areas. Tight schedules increase the requirement to provide detailed heavy lift analysis. The planning of every aspect of a critical lift operation is essential. Managing the behavior and trajectory of the lifted object during the lift is often left to the field crew. The rigger signalman and the crane operator communicate by radio, or by hand signals, to maneuver the lifted object between obstructions. This thesis presents advancements in the development of mathematical algorithms for the lift object trajectory path and analysis. The proposed methodology is divided into smaller manageable phases to control the process and at the same time create independent modules. Each step of the lifted object movement was algebraically-digitally tracked, starting at the lifted object pick-point through an optimum path development to the object’s final position or set-point. Parameters such as the minimum distance between the lifted object and passing obstructions and the minimum clearance between the lifted object and the crane boom envelope are some of the many predefined rules that were taken into account. Each step in the developed algorithm provides a short description, partial decision flowchart, and graphical interpretation of the problem, and some sections cover mathematical calculations of a defined path. The lifted object’s spatial trajectory analysis and optimization are part of the complex assignment relating to the crane selection process. The proposed methodology is tested on a case study, which is also described in this thesis in order to illustrate the essential features of the proposed methodology.

Using a naturalistic, longitudinal study design, 219 patients receiving treatment for headache disorders completed psychosocial assessments and 30-day daily diaries that assessed headache frequency, severity, and disability at pre-treatment and provided data on headache-related disability at pre-treatment and 1-, 2-, and 6-month follow-up. Latentclass trajectory analysis of the headache disability measure identified three treatment trajectory groups: (1) a high-disability non-responder group; (2) a high-disability responder group; and (3) a low-disability responder group.