College of Engineering Theses and Dissertations

Permanent URI for this collectionhttps://hdl.handle.net/1969.6/94188

Browse

Now showing 1 - 20 of 54

Evaluation of different GNSS solutions and SFM software workflows for surveying shorelines and remote areas using UAS
(2022-05) Pilartes-Congo Jr, José A.; Starek, Michael J.; Chu, Tianxing; Huang, Yuxia
The emergence and modernization of Unoccupied Aircraft Systems (UAS), broadly known as drones, and Structure-from-Motion (SfM) photogrammetry have made significant contributions to the geospatial and surveying world. Traditionally, indirect georeferencing by using ground control points (GCPs) is used to georeference UAS imagery when high accuracy positioning is required. However, this approach is tedious and impractical when surveying remote or inaccessible coastal areas, or when desiring to map coastlines from shipborne UAS operations. The broad applicability of UAS and SfM technologies has led to a wide range of data collection and SfM processing workflows that can be utilized, enhanced further by the implementation of various Global Navigation Satellite Systems (GNSS) techniques for direct georeferencing of the imagery. As part of an investigation conducted by the Office of Coast Survey (OCS) at the National Oceanic and Atmospheric Administration (NOAA), this study seeks to identify UAS-SfM data collection and processing workflows that maintain vertical accuracies at the decimeter-level without the aiding of GCPs. The study uses UAS imagery collected from two different UAS platforms at two different sandy beach study sites along the Southern Texas Gulf Coast. The objectives of the study are two-fold: (i) examine the applicability of Real-Time Kinematic (RTK), Post-Processed Kinematic (PPK), and Precise Point Positioning (PPP) GNSS solutions as plausible substitutes to ground control points (GCPs) for UAS-SfM shoreline mapping, and (ii) to evaluate the impact of three-commercial SfM software (Drone2Map, Metashape, and Pix4D) and one open-source software (Web OpenDroneMap) on the quantitative and qualitative characteristics of resulting mapping products. Results showed that RTK and PPK can reach centimeter-level vertical accuracies, fulfill the requirements set forth for this project, and are the most suitable alternatives to GCPs for remote surveying when plausible. When using PPK, the highest accuracies were reached when using base stations within 30 kilometers of the survey site, especially when combined with higher percentages of PPK fix, a measure that explains the number of photos that successfully underwent PPK correction. PPP offers the best alternative for remote UAS surveying, given that it is a single - receiver method, but the results evaluated here did not meet desired vertical accuracy levels. However, enhancing convergence time techniques is likely to reach even better results. In terms of SfM software, Metashape and Pix4D proved to be the most robust software altern atives achieving repeatable centimeter-level vertical accuracies for derived mapping products. Several inconsistencies were observed with Drone2Map and ODM, which hinder its applicability for UAS surveying without GCPs. The results and techniques discussed in this study help to optimize data acquisition and processing workflows for shoreline mapping and remote surveying.
The effects of neighborhood socioeconomic characteristics and spatial access to healthcare facilities on potentially preventable conditions of emergency department visits for adults in Texas coastal bend urban areas
(2019-05) Yang, Mei; Huang, Yuxia (Lucy); Jin, Lei; Zhao, Meng
The objective of this thesis is to explore the associate effects of spatial access to healthcare facilities and neighborhood socioeconomic characteristics on potentially preventable conditions of emergency department visits (PPCEDs) rates for adults ages 18 to 64 years old in the Coastal Bend urban areas of Texas. The emergency department (ED) visits data were obtained from all local hospital systems in the 15 counties in the Coastal Bend Area from September 1, 2009 to August 31, 2015. The neighborhood socioeconomic characteristics data, which include Education, Employment, Hispanic, Insurance, Language and Poverty, were obtained from the U.S. Census American Community Study 2010-2014 summary data. The adult PPCEDs were estimated based on the NYU Algorithm from ED visits data. The healthcare facilities data were obtained from InfoUSA, which is a residential and business database. The spatial access to hospitals and to general facilities, respectively, at the zip code level were measured by the enhanced two-step floating catchment area (E2SFCA) method. Both driving and transit models were considered. Multivariable regression models were used to estimate associations of adult PPCEDs rates with the spatial accessibility and socioeconomic factors. The main findings of spatial access (both driving and transit) to healthcare facilities are: 1) In the urban areas, Nueces county belongs to the highest spatial accessibility group; San Patricio county belongs to the second highest spatial accessibility group; other urban areas belong to the least spatial accessibility group with different values. 2) In all the Coastal Bend area, Nueces county (urban area) belongs to the highest spatial access group to healthcare facilities; however, the central part of San Patricio (rural area) belongs to the second highest spatial access group to hospitals; meanwhile, the central part of San Patricio, the north area of Kleberg and Brooks, and the south area of Jim Wells (rural areas) belong to the second highest spatial access group to general facilities with different values. For the neighborhood socioeconomic effects on adult PPCEDs rate in the Coastal Bend urban areas, 1) The neighborhood socioeconomic factors, Education and Insurance, have significantly positive correlations with adult PPCEDs rate. A higher rate of the adult population with a lower-than-high-school education level and a higher rate of the adult population in uninsured health situations resulted in a higher rate of adult PPCEDs at the zip code level. 2) However, Employee, Hispanic (racial/ethnic) and Poverty have negative but not significant correlations with PPCEDs rate at the zip code level. The increasing rate of the adult population of unemployed status, the total adult Hispanic population, and the adult population below-poverty-level decreased the PPCEDs rate. For the spatial accessibility effects on adult PPCEDs rate in the Coastal Bend urban areas, 1) Spatial access to hospitals and spatial access to general facilities have inverse effects on adult PPCEDs rate at the zip code level. Poorer access to hospitals and easier access to general facilities in urban areas contributed to a higher adult PPCEDs rate at the zip code level. The research results provide useful information for health providers and policy makers to take actions to increase equitable spatial access to healthcare facilities and increase the number of adults with medical home care. These actions could help decrease the PPCEDs use and improve the healthcare quality for adults in the Coastal Bend urban areas.
Detecting plant phenotypes from 3D point cloud data
(2019-08) Dani, Jimmy; King, Scott A.; Jung, Jinha; Belkhouche, Mohammed; Mandadi, Kranthi
In recent years, with the rapid development in indoor plant genotyping, there is a growing need for precise quantification of plant phenotypes. Currently, manual plant phenotyping is being used which is laborious, time-consuming, and prone to errors. This served as a motivation to develop an automated greenhouse phenotyping framework, that uses a 3D point cloud generated from RGB images. This study is focused on variations in plant phenotypes on different genotypes namely, Atlantic and Olalla, under controlled and drought stress treatment, throughout the growing season. The phenotypes considered in this study are: plant height, plant volume, leaf angle distribution and Excessive Greenness Index. Images of the plant are taken from two cameras hung on a post and a 3D point cloud is generated from those images. The phenotypes derived from the point cloud showed high correlation with manual measurements, which shows the system could be used for a variety of indoor plant phenotyping. The 99 percentile height shows the highest correlation with manually estimated height, and the volume and excessive greenness index results shows the Olalla genotype is more susceptible to stress as compared to the Atlantic genotype, the leaf angle distribution shows higher wilting for the drought stress treatment as compared to the control treatment.
Using a USV to efficiently clear uncertainty from aerial images for rescue boat path planning in flooded urban environments
(2019-05) Ozkan, Mehmet; King, Scott A.; Carrillo, Luis Garcia; Xie, Junfei
Mapping and path planning in disaster scenarios is an area that has bene ted from aerial imaging and unmanned aerial vehicles (UAVs). However, the integration of an unmanned surface vehicle (USV) in flood rescue operations has not received much attention. We propose a novel map generation and path planning algorithm, which makes use of aerial imaging provided by a UAV in combination with surface level information provided by a USV. Since the aerial image is a 2D projection of a 3D world, some areas of interest could be uncertain, such as under trees. Despite this issue, a Probabilistic Roadmap (PRM) path planning algorithm can be applied to the image in order to nd near-optimal paths for a rescue boat between initial and target locations. With the method proposed here, the preliminary PRM solution is further improved by means of an online feedback structure, where local information provided by the USV is incorporated to the overall map as soon as it becomes available, eliminating uncertainties. Simulation results demonstrate the effectiveness of the proposed approach for improving the time response in search and rescue operations in flooded areas.
Automated radar heading calibration with collaborating participants and multi-sensor fusion
(2021-12) Boyd, Josh; King, Scott A.; Li, Longzhuang; Wang, Wenlu
As unmanned aerial systems (UAS) become more prolific so will the use of radar systems for tracking UAS in the national airspace system (NAS). The future of Urban Air Mobility (UAM) involves large amounts of UAS operating autonomously and simultaneously in urban environments for the purpose of passenger or cargo transportation. Radar detection of UAS in an urban environment can be hindered by line of sight (LOS) blockage by large buildings thus necessitating many surveillance devices to gain full coverage. Currently, UAM is still in development by the Federal Aviation Administration (FAA) and other airspace partners and many cities do not have the need or resources for full radar coverage. Due to the high cost of individual radar systems and the quantity needed to cover urban areas it is currently not practical to have full radar coverage of an area at all times. Permanent stationary radar systems are generally calibrated once with occasional adjustments and low time constraints. Temporary radar systems must be calibrated and aligned before each mission deployment and often under short time constraints. Temporarily stationed mobile radar platforms will be utilized for specific targeted mission objectives until a more permanent solution is developed and implemented. In the case of disaster response or search and rescue, a temporary radar system needs to be quickly deployed. The key abilities required by a temporary radar system are accurate track position reporting and quick setup and breakdown. One of the bottlenecks to quick setup is heading calibration. Radar antenna alignment is crucial to the performance of the system and its ability to accurately determine the position of a tracked object. In this paper, we implement and compare multiple methods of radar heading calibration for accuracy and speed including manually with a handheld compass, manually with a web based heading helper tool, manually with a custom dual Real Time Kinetic (RTK) GPS alignment tool, and automated with a collaborating Radar Cross Section (RCS) device. For RCS devices we use a marine radar reflector and attached RTK GPS when unable to fly and an unmanned aerial vehicle (UAV) also with RTK GPS when able to fly. By leveraging our experience working with UAVs and Radars we show a method to autocalibrate the positioning sensors by using multisensory fusion and collaborating participants, thus reducing the amount of setup time, and increasing the accuracy of the system.
Intelligent mobile edge computing
(2021-12) Ale, Laha; King, Scott A.; Zhang, Ning; Li, Longzhuang; Hunag, Lucy; Rote, Carey
With the emergence of the Internet of Things (IoT), connected devices have been growing exponentially. These IoT devices typically have low resources, thus augmented resources, particularly compute and storage resources, are needed to support various IoT services. Mobile Edge Computing (MEC) deploys compute and storage resources at the network edge servers to accommodate IoT services, where data collected by IoT devices can be processed and analyzed in proximity. Compared with conventional cloud computing, MEC can mitigate potential network congestion caused by massive data transmission and reduce service latency. However, the performance of the MEC heavily relies on the prediction accuracy of the spatiotemporal distribution of IoT traffic and intelligent resource provision. In this work, we first developed a spatiotemporal method for modeling and predicting time-varying demand from IoTs so that MEC providers can provision resources efficiently. The prediction results can help network providers find the best suitable locations to deploy edge servers. Furthermore, we develop deep learning (DL) models to learn and predict the temporal content popularity to intelligently utilize the storage resources of MEC servers for caching content. Finally, deep reinforcement learning (DRL) models have been harnessed to control computational offloading to efficiently utilize computational resources to support IoT services and reduce energy consumption. The developed models are evaluated through simulations and real-world datasets, and the results show that our models outperform existing methods.
UAS mapping for oil spill response in sandy beach environments: Feasibility and best practices
(2021-12) Berryhill, Jacob; Starek, Michael J.; Chu, Tianxing; Gibeaut, Jim
Oil spill events can be catastrophically harmful to coastal ecosystems, causing considerable and long-term environmental and economic impacts before associated consequences are finally eliminated. Conducting timely, flexible, and accurate surveys immediately after a spill incident is of crucial importance for oil spill response in order to locate the spill, determine the size and volume of the spill, monitor and track the oil movement. Traditional survey methods and visual observations are usually performed for investigating the affected shoreline after an oil spill. Field sketches are used to record and convey the state of oiling in the affected areas. Diagnosis of oiling extent is limited to line-of-sight observations on the ground or by expensive manned aircraft operations. Recently, Unmanned Aircraft Systems (UAS) have been increasingly employed in various real-world applications, spanning from military scouting and scientific research to urban planning and entertainment. With the rapid development of miniaturized imaging and positioning technologies, UAS Structure-from-Motion (SfM) photogrammetry has become an emerging, cost-effective, and flexible solution for fulfilling various surveying and mapping needs at local scales. This thesis examined the potential and feasibility of using commercially available rotor copter and fixed-wing UAS platforms with SfM photogrammetric techniques to measure and monitor changes in beach elevation for shoreline oiling surveys. The state of the art of UAS-SfM together with its benefits and generic workflow in oil spill surveying were reviewed. A typical stretch of beach in South Texas was chosen as the study area in the thesis due to abundant historical data collected by the research laboratory from prior projects. The study site contains jetty blocks that provide stable features, as well as beaches that are both maintained and unmaintained. Four objectives were outlined with an effort to develop guidelines for UAS-SfM best practices for Shoreline Cleanup and Assessment Technique (SCAT). Based on the data collected at the study area, research findings suggest that without ground control points (GCPs), SfM processing with post-processing kinematic (PPK)-enabled image locations can achieve remarkably higher accuracy than that with autonomous Global Navigation Satellite System (GNSS) image geotags. Adding more GCPs can exponentially improve the overall accuracy for autonomous GNSS geotagged images. For the specific study area, the accuracy performance of the autonomous GNSS geotagged SfM products is on par with that of the differentially corrected GNSS geotagged SfM products with 10 GCPs used for georeferencing. By comparing against coordinates of the check points, the z residuals of a SfM-generated DSM were found better near the center of the beach and worse towards the water and in the dunes and vegetation. Another benefit of using a rigorous GCP control network is it significantly alleviates the bowling effect. Alternative solution for effectively alleviating the bowling effect in time critical survey missions where surveying GCPs is impossible is the use of high overlap oblique imagery and/or multi-elevation coverage. Height adjustments on erratic height values that occurred within autonomous GNSS geotagged images will not improve the accuracy in DSM rendering, however it is still recommended for reducing time and effort in identifying aerial target locations within the image set. When using the PPK operation mode, special attention needs to be paid because a consistent vertical datum should be maintained for the coordinates of both GCPs and image locations throughout the project. In case rapid SfM processing is considered essential for the sake of time, commercial SfM software demonstrated that several hours may be saved in terms of processing, but overall data quality of the geospatial products may have to be compromised.
Applications of deep learning and multi-perspective 2D/3D imaging streams for remote terrain characterization of coastal environments
(2021-12) Pashaei, Mohammad; Starek, Michael J.; Tissot, Philippe; King, Scott A.; Glennie, Craig L.; Lynch-Davis, Kathleen
Threats from storms, sea encroachment, and growing population demands put coastal communities at the forefront of engineering and scientific efforts to reduce vulnerabilities for their long-term prosperity. Updated and accurate geospatial information about land cover and elevation (topography) is necessary to monitor and assess the vulnerability of natural and built infrastructure within coastal zones. Advancements in remote sensing (RS) and autonomous systems extend surveying and sensing capabilities to difficult environments, enabling more geospatial data acquisition flexibility, higher spatial resolutions, and allowing humans to “see” in ways previously unattainable. Recent years have witnessed enormous growth in the application of small, unmanned aircraft systems (UASs) equipped with digital cameras for hyperspatial resolution imaging and dense three-dimensional (3D) mapping using structure-from-motion (SfM) photogrammetry techniques. In contrast to photogrammetry, light detection and ranging (lidar) is an active RS technique that uses a pulsed laser mounted on a static or mobile platform (from air or land) to scan in high definition the 3D structure of a scene. Rapid proliferation in lidar technology has resulted in new scanning and imaging modalities with ever increasing capabilities such as geodetic-grade terrestrial laser scanning (TLS) with ranging distances of up to several kilometers from a static tripod. TLS enables 3D sampling of the vertical structure of occluding objects, such as vegetation, and underlying topography. Full waveform (FW) lidar systems have led to a significant increase in the level of information extracted from a backscattered laser signal returned from a scattering object. With this technological advance and increase in remote sensing capabilities and data resolution, comes an increase in information gain at the cost of highly more complex and challenging big data sets to process and extract meaningful information. In this regard, utilizing end-to-end analyzing techniques recently developed in artificial intelligence (AI), in particular, convolutional neural network (CNN), developed under deep learning (DL) framework, seems applicable. DL techniques have recently outperformed state-of-the-art analysis techniques in a wide range of applications including RS. This work presents the application of DL for efficient exploitation of hyperspatial UAS-SfM photogrammetry and FW TLS data for land cover monitoring and topographic mapping in a coastal zone. Hyperspatial UAS images and TLS point cloud data with additional information about the scattering properties of illuminated target in the footprint of the laser beam encoded in returned waveform signals provide valuable geospatial data resources to uncover the accurate 3D structure of the surveyed environment. This study presents three main contributions: 1) Evaluation of different DCNN architectures, and their efficiencies, to classify land cover within a complex wetland setting using UAS imagery is investigated; 2) DCNN-based single image super-resolution (SISR) is employed as a pre-processing technique on low-resolution UAS images to predict higher resolution images over coastal terrain with natural and built land cover, and its effectiveness for enhancing dense 3D scene reconstruction with SfM photogrammetry is tested; 3) Full waveform TLS data is employed for point cloud classification and ground surface detection in vegetation using a developed DCNN framework that works directly off of the raw, digitized echo waveforms. Results show that returned raw waveform signals carry more information about a target’s spatial and radiometric properties in the footprint of the laser beam compared to waveform attributes derived from traditional waveform processing techniques. Collectively, this study demonstrates useful information retrieval from hyperspatial resolution 2D/3D RS data streams in a DL analysis framework.
Real-time human movement prediction based on human habits and emotion
(2021-08) Rodriguez, Roberto; Li, Longzhuang; Kar, Dulal; Pan, Chen
This research project develops a new deep neural network model for real-time human movement prediction based on human habits and emotion through the layering of neural networks, computer vision algorithms, and mathematical matching with a dynamic database. By combining multiple residual neural networks using different layering algorithms, the new model can increase prediction accuracies with reduced errors because of how each neural network adjusts to fill in the gaps left behind be each other to average out a proper evaluation of human motion. Specifically, the new model contains the following components: ResNet50 and altered ResNet34 on ImageNet for motion in all three dimensions, FURIA algorithm on Dlib 68 facial landmarks for emotion, and a nearest neighbor neural network for prediction based off locomotion. The result of this combination was that the prediction would take 0.05 milliseconds after emotion is selected to initiate its prediction path. The prediction module follows motion at first but with reduced accuracy until emotion is given to give a proper prediction with visualized projection. As its database of motions grows, the accuracy grows as well which leads to near real-time movement prediction. In the experiments, the new model outperforms existing models in both prediction accuracy and training speed due to this dynamic database. While other models require retraining of the neural networks to adjust to new testing data, this model relies on only database motion additions which greatly speeds up overall training and testing. The possible usages of this model are, but not limited to, health caregiving, culprit movements during police engagements, and child safety monitoring.
Fog prediction using deep learning models
(2021-05) Dinh, Hue Thi Hong; King, Scott A.; Tissot, Philippe; Li, Longzhuang; Collins, Waylon
The occurrence of fog has adverse impacts to human activities, aviation and water transportation operations. These events can cause postponement and cancellation of flights and accidents between ships and vessels leading to economic costs. The design of accurate models is required to forecast the low visibility events caused by fog. However, the prediction of fog remains a challenge due to the rare occurrence of these events. In this study, a deep learning networks (DNN) was proposed using the output from Numerical Weather Prediction (NWP) to predict 6-hour, 12-hour, and 24-hour lead time low visibility levels in the Corpus Christi Area. The model based on the autoencoder architecture was applied as a post-processing of deterministic NWP model and sea surface temperature (SST) output. The autoencoder was utilized to reduce the dimension of the input features to select a higher order of representation from raw data. By converting data from high dimensional space into a lower dimension, autoencoder models preserve the meaningful properties of original features in an unsupervised learning fashion. A logistic regression was also added to solve the classification problem of visibility level. Additionally, the under sampling and oversampling were also examined to solve the class imbalance problem caused by the less positive cases (fog cases). A 11-year database of NWP and SST was used to develop, train, validate, and test the proposed models to predict the occurrence of fog. The target of the models was categorized into three overlapping classes, including ≤ 1600m, ≤ 3200m, and ≤ 6400m. The prediction skill of these models was evaluated by relative operating characteristic curves and seven different skill scores. The results indicate that the DNN models are able to generate good discrimination for all lead times and visibility categories. The DNN models consistently outperformed an operational NWP model ensemble used by National Weather Service with respect to five skill scores (HSS, PSS, POD, CSI, and ORSS). The performance of the proposed models in dimensional reduction exceeds that of the combination between principal component analysis and logistic regression.
FSDCNN: A few shot detection mechanism that preserves its supervised nature
(2021-05) Agarwala, Mayank; King, Scott A.; Huang, Minhua; Li, Longzhuang
Object detection has become better with the advent of deep convolution neutral networks. However, the challenge of training a fully supervised system when there is a small amount of training samples available still remains. Another issue with fully supervised systems is seen upon encountering novel classes. It is difficult to retrain the model as it is a time-consuming and tedious process. Inspired by a human’s ability to learn at a rapid rate, few-shot learning models have seen rapid development. In contrast to fully supervised systems, these learn from just a few samples. We propose a few-shot detection model, FSDCNN, based on a two-stage detector, that optimizes both the region proposal network and the object detector with the help of few-shot learning. FSDCNN performs similar to other models when only 1 or 3 new samples are seen but outperforms them when 5 or 10 samples of the new classes are seen, and it preserves the fully supervised nature of the base two stage detector.
Few-shot learning with background subtraction
(2020-12) Wang, Hui; Li, Longzhuang; Zhang, Ning; Kar, Dulal
Few-shot learning for image classification aims to classify the image by only using few images as supporting samples. In the past several years, few-shot learning has achieved a huge improvement in image classification. In the recent work, such as meta-transferlearning (MTL) and Few-shot Adaptive Faster R-CNN have achieved a higher accuracy. In this paper, we are trying to combine three different methods together which are YOLOV2 model, Mask RCNN and our fewshot learning model. When a CNN wants to recognize animals in photos, there is a huge chance that even features that are supposed to represent trees will be encoded as belonging to those animals. Our main idea is to using YOLO algorithm, Mask RCNN and Opencv functions to reduce the noise and background as much as possible and keep our main object as it is. We would like to train our model using the image that only contain the object itself. We show that this approach is helpful to improve the accuracy in few-shot learning image classification.
Automatic canopy plot boundary detection using computer vision
(2020-05) Wynn, De Kwaan; King, Scott A.; Gonzales, Xavier; Li, Longzhuang
In Corpus Christi, Texas the United States Department of Agricultre (USDA) funded Texas A&M Agrilife research on large farmlands with hundreds of individual cotton vegetation plots. Each plot is planted uniformly in rows but not all plots grow at the same rate. Every week the plots are photographed using an Unmanned Aircraft System (UAS) flying at a height of 100 feet to record and evaluate growth for various reasons. The research scientists and farmer’s current method of localizing individual plots of vegetation within an image has proven to be very time consuming and inefficient. The algorithm developed in this paper automates the localization process under sunny conditions. The algorithm uses the Hue, Saturation, and Value (HSV) color space during the preprocessing stage to provide a binary image that indicates where each green pixel is located. Various OpenCV functions are then used to automate the crop localization process. Minimum and Maximum threshold values are set for filtering by size sections of the algorithm. Then, morphological operations are employed to further refine the regions of interest. The Connected Components function is used to determine how large each remaining object is and that size is then used to determine how large each localizing polygon will be drawn. After the size of each object is found, the size of each localizing polygon to be drawn is evaluated and split into smaller polygons whenever necessary. Not only the developed algorithm able to detect and classify cotton crop locations quickly but it is able to handle various complex situations. The developed method was evaluated by its accuracy, precision, and recall, which were 92.4%, 100%, and 92.4% respectively.
Evaluation of environmental impacts produced by gold mining areas on the surrounding forest in southwestern Ecuador using multispectral satellite and uas imagery
(2020-08) Veloz, Edison; Starek, Michael J.; Chu, Tianxing; Devlin, Donna
Mining is a dangerous activity that can cause environmental damage to flora and fauna due to the utilization of heavy metals. Ecuador has a long history of mineral extractions and nowadays the activity is increasing in many parts of the country. Environmentalists state that chemicals, such as cyanide and mercury, could cause alterations in vegetation health. This study utilizes satellite and Unmanned Aircraft System (UAS) based remote sensing to analyze impacts to vegetation health around a mining area located in Bella Rica within the El Oro province of the southwestern zone of Ecuador. Vegetation can be analyzed and identified through many remote sensing techniques, one of them is the Normalized Difference Vegetation Index (NDVI). This band ratio index ranges from +1 to -1 and uses red and near-infrared (NIR) bands to identify the presence of healthy or stressed vegetation. In this study, a small rotary UAS equipped with a two-band sensor recording red and NIR reflectance and a separate red-green-blue (RGB) digital camera was used to gather data and determine if vegetation closer to the mine exhibited different NDVI patterns compared to vegetation located farther away. Spatial differences in NDVI patterns may indicate potential impacts of waste from mining operations. To provide a time series assessment of vegetation changes around the mine, satellite imagery from PlanetScope was acquired and analyzed to measure changes in NDVI throughout the years 2017, 2018, and 2019. PlanetScope uses an array of miniaturized satellites, called CubeSats, equipped with four-band multispectral sensors providing imagery at a resolution of 3 m ground sample distance (GSD). In comparison, spatial resolution of the UAS products, which is dependent on flying height, range from 2.97 cm GSD for the RGB camera to 11.4 cm GSD for the multispectral sensor. Satellite derived NDVI was vi statistically compared to UAS derived NDVI values to assess the impact of spatial resolution and sensor quality on NDVI measurement. Furthermore, the UAS acquired RGB imagery was processed using Structure from Motion (SfM) photogrammetry to derive a 3D reconstruction of the scene, referred to as a point cloud. Properties of the point cloud data were analyzed to determine if relationships exist between land cover structure and NDVI patterns captured in the UAS multispectral imagery. From UAS based multispectral data, significant differences in NDVI values were found between vegetation close to the mining area and vegetation at longer distances (p < 0.05), indicating that mining waste could be altering NDVI values in the region. Satellite imagery analysis suggests that changes in NDVI are related to different human activities that have been developed inside the study area. UAS derived NDVI shows a strong linear relationship with PlanetScope derived NDVI (R = 0.91), suggesting that the low cost and light-weight sensor onboard the UAS was able to capture similar reflectance information but at much higher resolution. UAS-SfM point cloud data was applied to measure spatial variation in point density and canopy height, and determine if these measures could serve as a proxy for NDVI to assess vegetation health impacts from the mining operation. Results varied with NDVI and point cloud density exhibiting a weak relationship (R = 0.04). This relationship held at multiple resolutions suggesting that scene texture and uniformity in the densification stage of SfM does not correlate well with variation in NDVI due to differences in canopy cover. Interestingly, point cloud density changes did show a connection to the type of vegetation with high values of point density occurring over the more densely canopied forest areas. In contrast to point cloud density, UASSfM derived canopy height measures exhibited much stronger correlation to the UAS multispectral NDVI values (R = 0.69).
QoS driven optimal mobile edge server placement in mobile edge cloud
(2020-05) Sun, Peidong; Zhang, Ning; Li, Longzhuang; Carrillo, Luis
Mobile edge cloud is an emerging technology to enhance the Quality of Service (QoS) for mobile users' applications, especially for computation resource-consuming applications. A challenge in mobile edge cloud is the problem of mobile edge server placement, which concerns where to place the mobile edge servers to reduce the transmission delay and computation delay for tasks generated by mobile device users. In this essay, we work on the deployment of edge servers in the mobile edge cloud. We formulate the deployment problem as an integer linear problem and propose a density-based deployment of the MECs algorithm, which combines the K-means approach and integer linear programming. To evaluate the performance of our proposed approach, we conduct experiments using the telecom dataset of Shanghai. From the result of the experiment, we demonstrate that our proposed method could reduce the delay per task by around 10%.
Quadcopter pid controller design and path planning using bio-inspired meta-heuristic algorithms
(2020-05) Maddi, Dheeraj Reddy; Mahdy, Ahmed; Sheta, Alaa A.; Yadav, Mamta
The usage of Quadcopter in commercial fields has evolved significantly due to its phenomenal development. However, controlling the movements of a Quadcopter is a demanding task due to its complex dynamics. The usage of the Proportional-Integral-Derivative(PID) controller for stability control is quite challenging in regards to the complexity of Quadcopter’s nonlinear structure. Conventional methods like Ziegler-Nichols(ZN) for tuning the PID controller for a Quadcopter do not provide efficient performance and might also cause the system to be severely damaged. In this thesis, we are addressing the problem of the controlling a Quadcopter using Metaheuristic-based PID controller. Multi-Objective Fitness Function is proposed to reduce the overall time of the step response effectively. Path planning is one of the important concepts for a Quadcopter to move from one point to another point effectively. A novel Neighborhood Search Genetic Algorithm (NSGA) is presented for path planning by balancing the diversity inside the Genetic Algorithm using a Neighborhood Search to produce an efficient path. The performance of the NSGA has been compared to traditional A∗ , standard GA, and PSO. NSGA produced superior results in terms of cost.
A mixed reality testbed for design and validation of estimation and control strategies for unmanned systems research
(2020-08) Reyes, Gabriel; King, Scott A.; Garcia, Luis; Zhang, Ning
Unmanned Aircraft Systems (UAS) research and development groups validate their estimation and control strategies in simulations and real-life scenarios. However, various factors affect the time it takes to perform such tests as well as their quality, e.g. hardware malfunctions and physical limitations of the testing equipment and environment. This project focuses on generating and integrating virtual autonomous vehicles with real prototypes into a mixed reality space for evaluation of single and multi-agent UAS strategies. The calibration and alignment of the computer-generated world with respect to the inertial frame of the real world was accomplished based on the Umeyama algorithm. Making use of an entity existing in both real and virtual worlds, it was found that the pose error of the entity in the mixed reality environment with respect to the real world is always kept below 10%. A control strategy for enabling multiple virtual UAS for tracking a real ground vehicle was generated in the mixed reality environment. In our tests, a specific flight formation around the target vehicle was achieved by the virtual agents. Additionally, the flight formation around the moving target can be commanded online by a human user. Ultimately, the development of this project resulted in a modular solution that facilitates testing and evaluation of estimation and control strategies for autonomous agents, with the possibility of adaptation and expansion into different research domains.
Reef restoration facilitates habitat provisioning for oysters and motile epifauna
(2019-12) Martinez, Meghan Janessa; Pollack, Jennifer Beseres; Montagna, Paul; Withers, Kim
Severe degradation of oyster reef habitat over the past century has led to associated losses in ecological and economic benefits. Common oyster reef restoration goals target replacement of lost ecosystem services, including habitat provision, by replacing the ecological functions of lost reef habitats. The goal of this study was to monitor development of faunal communities on a restored oyster reef in the Gulf of Mexico. In July 2017, more than 1 M tons of reclaimed oyster shell were used to restore 1.83 ha of oyster reef complex (~610 linear m) in St. Charles Bay, Texas. Oysters, epifauna, and infauna were sampled monthly for the first three months after construction, and then were sampled quarterly for a total of 19 months at the restored reef and nearby reference sites. Within the first three months after construction, mean oyster densities increased by more than three times, growth rates peaked at 0.41 mm d-1 , and the restored oyster population shifted from 100 % spat to more than 90 % submarket size oysters. Although Perkinsus marinus infection was detected on every sampling date on the reference reef, only a single infected oyster was observed on the restored reef. Reef location—away from infected source populations— and other hydrological factors such as current speed and direction, may have impeded disease development. Epifaunal density, biomass, and diversity, became similar to that of the reference reef within four months after construction, but a shift in epifaunal community assemblages occurred between the first and the second year after construction, indicating monitoring periods of more than one year are necessary to capture faunal community development on a restored reef. The structure provided by the restored reef was conducive to oyster and epifaunal community development and may have supported ecological resistance since minimal impacts to reef structure were observed in the wake of Hurricane Harvey. Infaunal density, diversity, and biomass did not differ between sites adjacent (less than 5 m) versus distant (~30 m) from the restored reef and were governed more by salinity than presence of the restored reef. The recruitment and densities of oysters indicate that the restored reef met proposed success metrics within 19 months after construction, and that restored reefs can successfully replace ecosystem services, such as habitat provision, lost due to degradation.
Estimating sensorimotor disorders using Bayesian theory and probabilistic graphical models with mixed reality technology
(2019-12) Martinez, Juan; King, Scott A.; Baca, Jose; Murkherjee, M.
Assessing sensorimotor problems is challenging due to the level of uncertainty present in the Central Nervous System (CNS). To solve this problem, this work introduces a novel strategy that combines Bayesian Theory for motor control with Probabilistic Graphical Models to estimate sensorimotor disorders. Therefore, by integrating these frameworks, we provide a precise estimation of the presence, or absence, of a sensorimotor problem over time. As first step, the user performs a task in a Mixed Reality environment, which guidesthe user through a series of activitiesto examine their eye-hand coordination under uncertainty. This portion of the system collects sampled information about the user’s sensory output for analysis of the user’s optimal motor control. Thence, after collection of sensory information, the system detects patterns of drastic changes of variability by comparing the prior and posterior distributions of two senses (vision and proprioception). Then using statistical inference, we determine if the user is following a pattern of variability or not. To further support the fact of detecting a pattern of irregular control, we use a Bayesian Network to condition a user’s medical and personal information to infer their expected pattern of motor control variance. Then, we join the evidence from the observed and expected patterns to deduce an observed state. Ultimately, a sequence of observed states is provided to the Hidden Markov Model to estimate a sensorimotor disorder from the provided evidence. With the estimated results, from simulations, we obtained reliable information about the user’s sensorimotor performance overtime so proper decisions could be made to assess a user’s coordination condition. Therefore, the system is capable to estimate expected results regarding sensorimotor disorders when provided with multiple states of evidence.
Classification of medical Images using metaheuristic feature selection methods
(2019-12) Maddula, Kuladeep Anand Kumar; King, Scott A.; Sheta, Alaa A.; Yadav, Mamta
Magnetic Resonance Imaging (MRI) is a popular non-invasive diagnostic tool for brain imaging. Accurate analysis of brain MRI images help in early detection of brain tumors and could save lot of lives. But accurate classification of the images as normal or pathological is a challenging task from the clinical as well as technology stand point. Brain MRI images consists of a large information set which contain redundancy in determining the condition of the brain. The redundant information would lead to increase in dimensionality of the data. Therefore, using a feature selection algorithm to find an optimum set of features would reduce the time and computation complexity of the classifiers for distinguishing the brain MRI images. This work is to study the performance of feature selection with different meta-heuristic search algorithms with multiple fitness functions. The three meta-heuristic algorithms considered are Binary Genetic Algorithm, Binary Particle Swarm Optimization and Binary Grey Wolf Optimizer for selecting an optimal set of features out of the extracted features from brain MRI images. The feature selection is performed on the 13 statistical features extracted from the brain MRI images using Discrete Wavelet Transform, Principle Component Analysis and Grey Level Co-occurrence matrix. The performance of the feature selection algorithms are compared by applying 4 different sets of features from each algorithm to seven different test classifiers. Our results obtained show high performance using feature selection.

Browse

Recent Submissions