The determination of the 3D motion of a camera and the 3D structure of the scene in which the camera is moving, known as the Structure from Motion (SFM) problem, is a central problem in computer vision. Specifically, the recursive (online) estimation is of major interest for robotics applications such as navigation and mapping. Many problems still hinder the deployment of SFM in real-life applications namely, (1) the robustness to noise, outliers and ambiguous motions, (2) the numerical tractability with a large number of features and (3) the cases of rapidly varying camera velocities. Towards solving those problems, this research presents the following four contributions that can be used individually, together, or combined with other approaches. A motion-only filter is devised by capitalizing on algebraic threading constraints. This filter efficiently integrates information over multiple frames achieving a performance comparable to the best state of the art filters. However, unlike other filter based approaches, it is not affected by large baselines (displacement between camera centers). An approach is introduced to incorporate, with only a small computational overhead, a large number of frame-to-frame features (i.e., features that are matched only in pairs of consecutive frames) in any analytic filter. The computational overhead grows linearly with the number of added frame-to-frame features and the experimental results show an increased accuracy and consistency. A novel filtering approach scalable to accommodate a large number of features is proposed. This approach achieves both the scalability of the state of the art filter in scalability and the accuracy of the state of the art filter in accuracy. A solution to the problem of prediction over large baselines in monocular Bayesian filters is presented. This problem is due to the fact that a simple prediction, using constant velocity models for example, is not suitable for large baselines, and the projections of the 3D points that are in the state vector can not be used in the prediction due to the need of preserving the statistical independence of the prediction and update steps.

The Internet has profoundly changed todays human being life. A variety of information and online services are offered by various companies and organizations via the Internet. Although these services have substantially improved the quality of life, at the same time they have brought new challenges and difficulties. The information security can be easily tampered by many threats from attackers for different purposes. A catastrophe event can happen when a computer or a computer network is exposed to the Internet without any security protection and an attacker can compromise the computer or the network resources for destructive intention.<p> The security issues can be mitigated by setting up a firewall between the inside network and the outside world. A firewall is a software or hardware network device used to enforce the security policy to the inbound and outbound network traffic, either installed on a single host or a network gateway. A packet filtering firewall controls the header field in each network data packet based on its configuration and permits or denies the data passing thorough the network.<p> The objective of this thesis is to design a highly customizable hardware packet filtering firewall to be embedded on a network gateway. This firewall has the ability to process the data packets based on: source and destination TCP/UDP port number, source and destination IP address range, source MAC address and combination of source IP address and destination port number. It is capable of accepting configuration changes in real time. An Altera FPGA platform has been used for implementing and evaluating the network firewall.

In-process turbine monitoring has been a subject of research since the advent of gas turbines; however, it is difficult because it requires precision measurements to be made at high speeds and temperatures. The measurement of turbine blade tips is especially intriguing because of the potential it holds to greatly increase the efficiency of engine operation and maintenance. Tip-to-casing clearance is one of the major sources of inefficiency in a turbine and monitoring of this clearance would allow active tip-clearance control systems to be implemented. Also, analysis of engine wear through vibration monitoring may increase the effectiveness of engine maintenance and repair. A sensor recently developed at Georgia Tech could answer this challenge. The sensor operates by measuring the phase change of reflected microwaves to measure blade tip displacement. It is robust even in the harsh turbine environment. However, in sensor measurements, the microwave beam pattern causes a phenomenon called spatial filtering to occur, which may compromise the precision of measurements. Since the beam is not a thin line reflecting off a single point on the turbine blade, measurements are a weighted average of measurements along the entire surface within the field-of-view of the sensor. The net effect is a blurred measurement. In measuring turbine blades, only the tip is vital, so the blurring in between blades is not extremely detrimental. However, changing measurement geometry affects the amount of spatial filtering and hence the accuracy of the measurement. This thesis presents a detailed analysis of this phenomenon and especially its effect on turbine blade tip clearance measurements. A design of experiments is presented to qualitatively understand the effect of geometric factors on tip measurements. Along with experimentation, a robust, three-dimensional, ray-tracing, electromagnetic model is presented which was developed to further understand spatial filtering and to analyze specific geometric factors in the measurement of turbine blades. The research shows that microwave measurements may still be made to sufficient accuracy even considering the effect of spatial filtering, and by quantifying spatial filtering in measurements, it may be possible in to glean additional useful data from measurements.

Multiple Hypothesis Tracking (MHT) is generally the preferred data association technique for tracking targets in clutter and with missed detections due to its increased accuracy over conventional single-scan techniques such as Nearest Neighbor (NN) and Probabilistic Data Association (PDA). However, this improved accuracy comes at the price of greater complexity. Sequential K-best MHT is a simple implementation of MHT that attempts to achieve the accuracy of multiple hypothesis tracking with some of the simplicity of single-frame methods. Our first major objective is to determine under what general conditions Sequential K-best data association is preferable to Probabilistic Data Association. Both methods are implemented for a single-target, single-sensor scenario in two spatial dimensions. Using the track loss ratio as our primary performance metric, we compare the two methods under varying false alarm densities and missed-detection probabilities. Upon implementing a single-target Sequential K-best MHT tracker, a fundamental problem was observed in which the tracks coalesce. The second major thrust of this research is to compare different approaches to resolve this issue. Several methods to detect track coalescence, mostly based on the Mahalanobis and Kullback-Leibler distances, are presented and compared.

This thesis deals with the problem of tracking highly deformable objects in the presence of noise, clutter and occlusions. The contributions of this thesis are threefold: A novel technique is proposed to perform filtering on an infinite dimensional space of curves for the purpose of tracking deforming objects. The algorithm combines the advantages of particle filter and geometric active contours to track deformable objects in the presence of noise and clutter. Shape information is quite useful in tracking deformable objects, especially if the objects under consideration get partially occluded. A nonlinear technique to perform shape analysis, called kernelized locally linear embedding, is proposed. Furthermore, a new algebraic method is proposed to compute the pre-image of the projection in the context of kernel PCA. This is further utilized in a parametric method to perform segmentation of medical images in the kernel PCA basis. The above mentioned shape learning methods are then incorporated into a generalized tracking algorithm to provide dynamic shape prior for tracking highly deformable objects. The tracker can also model image information like intensity moments or the output of a feature detector and can handle vector-valued (color) images.

The thesis is mainly focused on a hybrid location system, which processes wireless and inertial measurements by extended Kalman filtering. Inertial location system is usually used with Dead-Reckoning method, which calculates the present location and heading direction from a previous known state by using measurements of accelerometer and gyroscope, which have immunity from the environment. The system estimates the position by integrates the measurements of sensors, resulting in high accuracy during a short period. However, the unreliability grows with time due to the bias effect on sensors. By combining the wireless location and inertial system, the uncertainty of estimation can be reduced. In wireless communications, the locations of base stations and the times of signal arrival can be used in locating a mobile station. However, signal propagation could be blocked by objects. The non-line of sight (NLOS) effects cause arrival delay and is usually modeled as exponential distributions. Previously, the improved biased Kalman filters were designed to mitigate the NLOS effect in base station measurements. The system design has difficulty in accommodating inertial measurements. The inertial has immunity to the environment. The property is of help in the NLOS mitigation. Therefore, we propose a hybrid location system that integrating the wireless and inertial measurements by using a hybrid biased extended Kalman filter at the stage of positioning. The system provides better prediction with the assistance of enviroment-free inertial measurements. The NLOS mitigation with prediction feedback scheme results in better mitigation performance. Simulations of different situations have been conducted based on parameters in the IEEE 802.15.3a ultra-wideband environment. The performance differences between the proposed method and other approaches show that inertial assisted system effectively reduces the NLOS effects. Also, the proposed hybrid location system has more efficient mitigation performance and better tracking results.

Traditional scheduling tools like Gantt Charts and CPM while useful in planning and execution of complex construction projects with multiple interdependent activities haven?t been of much help in implementing effective control systems for the same projects in case of deviation from their desired or assumed behavior. Further, in case of such deviations project managers in most cases make decisions which might be guided either by the prospects of short term gains or the intension of forcing the project to follow the original schedule or plan, inadvertently increasing the overall project cost. Many deterministic project control methods have been proposed by various researchers for calculating optimal resource schedules considering the time-cost as well as the time-cost-quality trade-off analysis. But the need is for a project control system which optimizes the effort or cost required for controlling the project by incorporating the stochastic dynamic nature of the construction-production process. Further, such a system must include a method for updating and revising the beliefs or models used for representing the dynamics of the project using the actual progress data of the project. This research develops such an optimal project control method using Kalman Filter forecasting method for updating and using the assumed project dynamics model for forecasting the Estimated Cost at Completion (EAC) and the Estimated Duration at Completion (EDAC) taking into account the inherent uncertainties in the project progress and progress measurements. The controller is then formulated for iteratively calculating the optimal resource allocation schedule that minimizes either the EAC or both the EAC and EDAC together using the evolutionary optimization algorithm Covariance Matrix Adaption Evolution Strategy (CMA-ES). The implementation of the developed framework is used with a hypothetical project and tested for its robustness in updating the assumed initial project dynamics model and yielding the optimal control policy considering some hypothetical cases of uncertainties in the project progress and progress measurements. Based on the tests and demonstrations firstly it is concluded that a project dynamics model based on the project Gantt chart for spatial interdependencies of sub-tasks with triangular progress rates is a good representation of a typical construction project; and secondly, it is shown that the use of CMA-ES in conjunction with the Kalman Filter estimation and forecasting method provides a robust framework that can be implemented for any kind of complex construction process for yielding the optimal control policies.

Parametric uncertainty is one of many possible causes of divergence for the Kalman filter. Frequently, state estimation errors caused by imperfect model parameters are reduced by including the uncertain parameters as states (i.e., augmenting the state vector). For many situations, this not only improves the state estimates, but also improves the accuracy and precision of the parameters themselves. Unfortunately, not all filters benefit from this augmentation due to computational restrictions or because the parameters are poorly observable. A parameter with low observability (e.g., a set of high order gravity coefficients, a set of camera offsets, lens calibration controls, etc.) may not acquire enough measurements along a particular trajectory to improve the parameter's accuracy, which can cause detrimental effects in the performance of the augmented filter. The problem is then how to reduce the dimension of the augmented state vector while minimizing information loss. This dissertation explored possible implementations of reduced-order filters which decrease computational loads while also minimizing state estimation errors. A theoretically rigorous approach using the ?consider" methodology was taken at discrete time intervals were explored for linear systems. The continuous dynamics, discretely measured (continuous-discrete) model was also expanded for use with nonlinear systems. Additional techniques for reduced-order filtering are presented including the use of additive process noise, an alternative consider derivation, and the minimum variance reduced-order filter. Multiple simulation examples are provided to help explain critical concepts. Finally, two hardware applications are also included to show the validity of the theory for real world applications. It was shown that the minimum variance consider Kalman filter (MVCKF) is the best reduced-order filter to date both theoretically and through hardware and software applications. The consider method of estimation provides a compromise between ignoring parameter error and completely accounting for it in a probabilistic sense. Based on multiple measures of optimality, the consider filtering framework can be used to account for parameter error without directly estimating any or all of the parameters. Furthermore, by accounting for the parameter error, the consider approach provides a rigorous path to improve state estimation through the reduction of both state estimation error and with a consistent variance estimate. While using the augmented state vector to estimate both states and parameters may further improve those estimates, the consider estimation framework is an attractive alternative for complex and computationally intensive systems, and provides a well justified path for parameter order reduction.

Images may be degraded by some random error which is called noise. Noise may occur during image capture, transmission or processing and its elimination is achieved by smoothing filters. Linear smoothing filters blur the edges and edges are important characteristics of images and must be preserved. Various edge preserving smoothing filters are proposed in the literature. In this thesis, most common smoothing and edge preserving smoothing filters are discussed and a new method is proposed by modifying Ambrosio Tortorelli approximation of Mumford Shah Model. New method takes into edge direction consistency account and produces sharper results at comparable scales.

none