Dr. Hakki Erhan Sevil

Traditional discussions of phase correlation for image registration consider the usefulness of the technique for feature-rich images and generally restrict image shifts to small percentages of the image size. When applied to featureless images, in applications such as cloud tracking, the phase correlation output is degraded by a prominent noise component, which is not predicted by the fundamental mathematical expression for phase correlation. In this work, an additional term is proposed in the mathematical description of input images, and expressions for the phase correlation under a variety of situations are developed in this manner. Based on the mathematical expressions, several image pre-processing approaches are proposed to improve phase correlation results for featureless imagery. A large set of sky images is used to represent the featureless image category and phase correlation results are analyzed for each proposed pre-processing technique and compared to the basic phase correlation algorithm. Results show dramatic enhancements in phase correlation results.

One of the most important challenges in an autonomous and robotics system is the path planning in which the system finds the optimal path from start point to goal point. The traditional path planning algorithms may have large memory requirements which scale with the size and resolution of the configuration space. To address these challenges, this paper introduces a novel path planning algorithm that combines Particle Swarm Optimization and Artificial Potential Field in the form of a path planning algorithm for mobile robots. The biological and physical concepts from Particle Swarm Optimization and Artificial Potential Field algorithms are combined to yield an algorithm which minimizes instances of getting stuck in local minima and generates a smooth but feasible path. The developed method requires memory which scales only with the number of particles and the time taken to reach the goal. This results in a memory-efficient solution that generates smooth and feasible paths for mobile robots navigating in a 2D space.

The Wumpus World scenario is an exploration into the application of artificial intelligence to navigate through a world with pits and a fictional character called the 'Wumpus'. The intent for the agent is to navigate through the world and find the gold without being mauled by the Wumpus or fall into a pit. The game is based exclusively on rules of behaviour upon perceptions at the agent's current location and previous locations stored in the agent's knowledge base. The intention of this paper is to present our developed knowledge base (KB)-based algorithms with implementation to the Wumpus World scenario to show proof-of-concept results. Using minimum remaining value and KB approach, single agent and multi-agent algorithms are developed and tested in simulation environment. In multi-agent scenario, the developed algorithm also constructs a common KB with perception inputs from different agents, and the common KB can be accessed by any agent in the system. As a part of larger intelligent robotic development effort, this implementation study seeks to take the traditional single agent book example and expand it to a multi-agent perspective. The simulation results show that our developed algorithms successfully performed in both single and multi-agent versions.

Natural disasters are a major source of significant damage and costly repairs around the world. After a natural disaster occurs, there is usually a significant amount of damage, and with that, there are also a lot of costs involved with repairing and aiding all the people involved. In addition, the occurrence of natural phenomena has increased significantly in the past decade. With that in mind, post-disaster damage detection is usually performed manually by human operators. Taking into consideration all the areas one has to closely look into, as well as the difficult terrain and places with hard access, it becomes easy to understand how incredibly difficult it is for a surveyor to identify and annotate every single possible damage out there. Because of that, it has become essential to find new creative solutions for damage detection and classification in the case of natural disasters, especially hurricanes. This study focuses on the feasibility of using a Visual Question Answering (VQA) method for post-disaster damage detection, using aerial footage taken from an Unmanned Aerial Vehicle (UAV). Two other approaches are also utilized to provide comparison and to evaluate the performance of VQA. Our case study on our custom dataset collected after Hurricane Sally shows successful results using VQA for post-disaster damage detection applications.

This research effort proposes a novel method for identifying and extracting roads from aerial images taken after a disaster using graph-based image segmentation. The dataset that is used consists of images taken by an Unmanned Aerial Vehicle (UAV) at the University of West Florida (UWF) after hurricane Sally. Ground truth masks were created for these images, which divide the image pixels into three categories: road, non-road, and uncertain. A specific pre-processing step was implemented, which used Catmull-Rom cubic interpolation to resize the image. Moreover, the Gaussian filter used in Efficient Graph-Based Image Segmentation is replaced with a median filter, and the color space is converted from RGB to HSV. The Efficient Graph-Based Image Segmentation is further modified by (i) changing the Moore pixel neighborhood to the Von Neumann pixel neighborhood, (ii) introducing a new adaptive isoperimetric quotient threshold function, (iii) changing the distance function used to create the graph edges, and (iv) changing the sorting algorithm so that the algorithm can run more effectively. Finally, a simple function to automatically compute the k (scale) parameter is added. A new post-processing heuristic is proposed for road extraction, and the Intersection over Union evaluation metric is used to quantify the road extraction performance. The proposed method maintains high performance on all of the images in the dataset and achieves an Intersection over Union (IoU) score, which is significantly higher than the score of a similar road extraction technique using K-means clustering.

The aim of this research effort was to develop a framework for a structure from motion (SfM)-based 3D reconstruction approach with a team of autonomous small unmanned aerial systems (sUASs) using a distributed behavior model. The framework is composed of two major goals to accomplish this: a distributed behavior model for a team of sUASs and a SfM-based 3D reconstruction using team of sUASs. The developed distributed behavior model is based on the entropy of the system, and when the entropy of the system is high, the sUASs get closer to reducing the overall entropy. This is called the grouping phase. If the entropy is less than the predefined threshold, then the sUASs switch to the 3D reconstruction phase. The novel part of the framework is that sUASs are only given the object of interest to reconstruct the 3D model, and they use the developed distributed behavior to coordinate their motion for that goal. A comprehensive parameter analysis was performed, and optimum sets of parameters were selected for each sub-system. Finally, optimum parameters for two sub-systems were combined in a simulation to demonstrate the framework's operability and evaluate the completeness and speed of the reconstructed model. The simulation results show that the framework operates successfully and is capable of generating complete models as desired, autonomously.

This study presents a novel distributed behavior model for multi-agent unmanned aerial vehicles (UAVs) based on the entropy of the system. In the developed distributed behavior model, when the entropy of the system is high, the UAVs get closer to reduce the overall entropy; this is called the grouping phase. If the entropy is less than the predefined threshold, then the UAVs switch to the mission phase and proceed to a global goal. Computer simulations are performed in AirSim, an open-source, cross-platform simulator. Comprehensive parameter analysis is performed, and parameters with the best results are implemented in multiple-waypoint navigation experiments. The results show the feasibility of the concept and the effectiveness of the distributed behavior model for multi-agent UAVs.

This paper outlines the methods, results, and statistical analysis of a model we developed to demonstrate the feasibility of applying remote sensor meteorological data to navigation by using meteorological contour matching (METCOM). Terrain contour matching (TERCOM), a contemporary navigation system, possesses inherent performance flaws that may be resolved and improved by METCOM for subsonic and hypersonic missile or aircraft navigation. Remote sensor imagery data for this model was accessed from the Geostationary Operational Environmental Satellites-R Series operated by the National Oceanic and Atmospheric Administration by using Amazon Web Services through a script we developed in Python. Data processed for the model included imagery data and corresponding geospatial data from the legacy atmospheric profile products: legacy vertical temperature and legacy vertical moisture. Our analysis of the model included an error assessment to determine model accuracy, geostatistical analysis through semivariograms, meteorological signal of model data, and a combinatorial analysis to evaluate navigation performance. We conducted a model assessment which indicated an accuracy of 66.2% in the data used as a combined result of instrument error and interference of cloud formations. Results of the remaining analysis offered methods to evaluate METCOM performance and compare different meteorological data products. These results allowed us to statistically compare METCOM and TERCOM, yielding several indications of improved performance including an increase by a factor of at least 13.5 in data variability and contourability. The analysis we conducted served as a proof of concept to justify further research into the feasibility and application of METCOM.

This study presents the implementation of basic nursing tasks and human subject tests with a mobile robotic platform (PR2) for hospital patients. The primary goal of this study is to define the requirements for a robotic nursing assistant platform. The overall designed application scenario consists of a PR2 robotic platform, a human subject as the patient, and a tablet for patient-robot communication. The PR2 robot understands the patient's request and performs the requested task by performing automated action steps. Two categories and three tasks are defined as: patient sitter tasks, include object fetching and temperature measurement, and patient walker tasks, including supporting the patient while they are using the walker. For this designed scenario and these tasks, human subject tests are performed with 27 volunteers in the Assistive Robotics Laboratory at the University of Texas at Arlington Research Institute (UTARI). Results and observations from human subject tests are provided. These activities are part of a larger effort to establish adaptive robotic nursing assistants (ARNA) for physical tasks in hospital environments.

In this study, we aimed to identify spatial clusters of countries with high rates of cyber attacks directed at other countries. The cyber attack dataset was obtained from Canadian Institute for Cybersecurity , with over 110,000 Uniform Resource Locators (URLs), which were classified into one of 5 categories: benign, phishing, malware, spam, or defacement. The disease surveillance software SaTScanTM was used to perform a spatial analysis of the country of origin for each cyber attack. It allowed the identification of spatial and space-time clusters of locations with unusually high counts or rates of cyber attacks. Number of internet users per country obtained from the 2016 CIAWorld Factbook was used as the population baseline for computing rates and Poisson analysis in SaTScanTM. The clusters were tested for significance with a Monte Carlo study within SaTScanTM, where any cluster with p < 0.05 was designated as a significant cyber attack cluster. Results using the rate of the different types of malicious URL cyber attacks are presented in this paper. This novel approach of studying cyber attacks from a spatial perspective provides an invaluable relative risk assessment for each type of cyber attack that originated from a particular country.