While various online multi-target tracking methods have been proposed recently, most of their runtime speed struggle at 1-10 frames per second for moderate crowded scenes. In this thesis, we present a novel real-time multi-target tracking system based on the tracking-by-detection framework. Our system is designed for tracking a variable number of interacting targets from a single, static, above shoulder camera, which is a general setting for video surveillance. One challenge in our approach is that when background subtraction is used for detecting moving targets, merged measurements occur frequently because of target interactions. To cope with the problem, we propose to use correlation filter based object detector to robustly separate the targets in merged measurements. Then, online object tracking assisted data association is used to solve the track-measurement assignment. To reduce computation load, our object tracking algorithm is assisted by correlations filter based trackers which share the same features used by our object detector. In addition, to recover partially occluded targets, we allow unconfident detections to be assigned to tracks whilst care is taken to avoid introducing additional false positives. We also analyze the online approximation to multi-channel correlation filters. Our experiments show that exact solution is more resistant to noisy channels than approximate solution. Evaluation on generally accepted datasets reveals that the proposed system is comparable to state-of-the-art methods in terms of performance while running several magnitudes faster. Additionally, we show that the proposed system can be readily implemented on the Texas Instruments TMS320C6678 DSP (C6678) without significant degradation in speed or performance. Details on efficient implementation of the system is also discussed. Especially, for computing Histogram of Oriented Gradients (HOG) feature, our optimized implementation runs at 60fps on VGA images on a single core of C6678, which is 10 times faster than a directly ported implementation. / published_or_final_version / Computer Science / Master / Master of Philosophy

American education is built upon a cultural paradigm of equality and access. Tracking students into homogenous classes based on prior academic performance could disadvantage lower-achieving students, thereby reinforcing inequality. The problem in the study district is that homogenous tracking was implemented, yet the system had not been evaluated within the context of implementation. The purpose of this qualitative case study was to examine educators' perceptions of the practice of tracking. Bandura and Maslow's social cognitive theory provided the conceptual framework and Dewey's beliefs on social justice and a spirit of equity provided the theoretical foundation. Open-ended narrative questionnaires were disseminated to approximately 109 educators in a public school district in rural northwestern New Jersey. Ten purposefully sampled interviews were also conducted for triangulation and to reach a robust understanding of the qualitative data. The data were content-analyzed through open coding and categorizing of emergent themes. The findings indicated a gap between existing district cultural norms and both current and seminal research as educators supported the district's practice of tracking. The majority of participants stated that creating homogenous classroom settings, based upon student behavior, work ethic, and motivation, improves the instructional environment for educators. The results informed the development of a white paper for the school board and district stakeholders with policy recommendations for the local tracking model. The implications for positive social change are that these endeavors may inspire the consideration of heterogeneous grouping models to better support student learning, student self-efficacy, and equity in how students are served.

A pellet target is an internal target system for accelerator experiments in nuclear and particle physics. The target consist of small spheres of frozen hydrogen, called pellets. The pellets interact with high energy accelerator-beam particles in a particle accelerator. The challenge is to track these pellets for good accuracy of interaction position in both time and space. The pellets are tracked with lasers and cameras. The main goal of this project was to develop a method to find the best time resolution and to optimize the efficiency of the pellet tracking system. This project addresses challenges with making trustworthy measurements were stability in the setup, difficulties with alignment and optimizing of exposure cycles. Because of stability issues, a more stable and robust module that also will ease adjustment of alignment is under construction. A well-aligned setup can be confirmed in two ways. Firstly by confirming that the cameras detect pellet signals for a about 300 micron-height change of the laseror the camera and secondly that a focus interval is within 100 micron. A mathematical model that calculates class-distributions for varying exposure cycles and shifts can be used to predict the results from a measurement with pellets. The model can be used for testing an appropriate exposure cycle before running it with pellets. In addition, an appropriate laser power should be chosen since the effects of both too low or too high power gives misleading results. This project has contributed to progress in the field of pellet tracking.

Location tracking is of paramount importance to many applications such as healthcare, retail and navigation. Outdoor tracking can be easily implemented using the Global Positioning System (GPS). However, indoor tracking has been a difficult problem to tackle because GPS requires the line of sight to the satellites and therefore it does not work well in indoor environments. In this thesis, a high-precision indoor tracking system is proposed to identify, locate and track a person in an indoor room at a low cost. The proposed tracking system consists of three components: StepscanTM tiles, RFID and Kalman-filter based prediction. The StepscanTM tiles can generate precise location information. However, using StepscanTM tiles only in an indoor tracking system is too expensive because the manufacturing cost of each StepscanTM tile is very high. In the proposed system, StepscanTM tiles are deployed to cover a part of the indoor floor while RFID provides a full coverage. The location information from StepscanTM tiles and RFID is then used as inputs for our innovative prediction algorithm based on the Kalman filter, which consequently generates high-precision tracking results. The performance of the proposed system is investigated through extensive simulations. Our simulation results indicate that the proposed system increases the capability to track and locate a person by at least 24% (more than 50% in some cases), with errors ranging from 2.5% to 15%. Furthermore, the proposed system helps to reduce the cost of indoor tracking significantly. In terms of the number of StepscanTM tiles deployed in the system, a reduction of 7 to 25 tiles can be achieved in the scenarios under investigation. In terms of monetary cost, $21,000 to $75,000 can be saved for an indoor tracking system considered in our research.

No description available.

A general overview of the hardware and software implementation requirements of an AC magnetic tracking system is presented. The overview includes an analysis of the Polhemus Fastrak, the development of mathematical models and algorithms used for magnetic tracking and the design of test jigs used to verify the results. An original non-iterative tracking algorithm that allows sensor position and orientation to be calculated in rotation matrix form is developed and verified. Future methods of calibration are proposed with results supporting the design methods. It is concluded that the algorithm presented is effective, but that further developments in hardware and software are needed in order for the system to attain the tracking resolution and accuracy specifications of other leading tracking systems.

Thesis (M.S.)--Ohio University, March, 1986. / Title from PDF t.p.

Thesis (Ph. D.)--University of Adelaide, Dept. of Electrical and Electronic Engineering, 1999. / Includes bibliographical references (leaves 106-110).

Thesis (M.S.)--University of Wisconsin--Madison, 1982. / Typescript. eContent provider-neutral record in process. Description based on print version record. Includes bibliographical references (leaves 106-107).

This study addresses the development of a methodology using the Doppler Effect for high-resolution, short-range tracking of small projectiles and vehicles. Minimal impact on the design of the moving object is achieved by incorporating only a transmitter in it and using ground stations for all other components. This is particularly useful for tracking objects such as sports balls that have configurations and materials that are not conducive to housing onboard instrumentation. The methodology developed here uses four or more receivers to monitor a constant frequency signal emitted by the object. Efficient and accurate schemes for filtering the raw signals, determining the instantaneous frequencies, time synching the frequencies from each receiver, smoothing the synced frequencies, determining the relative velocity and radius of the object and solving the nonlinear system of equations for object position in three dimensions as a function of time are developed and described here.