136879 research outputs found
Sort by
Improved Bi-Level Thresholding Based Iterative Analysis for Concrete Building-Surface Damage Detection in a Post-Earthquake Environment
No full textPost-earthquake impact assessment plays a critical role in optimal response planning and rescue resource allocation. Manual building damage inspection in a post-earthquake environment is typically resource-consuming, and prone to limitations based on subjectivity. Work in this field has benefited from the advancement of contact or non-contact sensing techniques, however, in a post-earthquake response phase, building inspection is still conducted using manual or semi-automated techniques. In this study, an improved bi-level thresholding-based iterative framework is proposed to automate the delineation of concrete-spalling using 3D point cloud data. Point curvature is used for damage point characterization, and two stopping conditions are defined for process automation. This study is focused on identifying the surface defects on a per-building level to delineate spalling defects on EMS-98 damaged level 3-4 structures.
The proposed algorithm is tested for damaged point labeling in synthetic data and data captured from a real-world scenario. Initially, synthetic element-level data was generated using CAD models and random noise generators were used to evaluate the performance of the proposed algorithm. Results from the analysis are compared with the state-of-the-art (SOA) iterative refinement analysis. Comparative analysis demonstrated that the proposed algorithm rendered damaged region detection with improved completeness and correctness. In this study, the use of Matthews Correlation Coefficient (MCC) and mean Generalized Intersection over Union(mGIoU) metrics for performance evaluation is proposed. For low-noise rectangular-wall samples used, an average increase of 20%and 55% MCC value compared to generalized iterative refinement analysis was observed for stopping condition-I & II respectively. Similarly, an average increase of 31% and 21% mGIoU is observed. Furthermore, the proposed algorithm demonstrated the potential to generalize better compared to the state-of-the-art as it does not require tuning or training using ground truth data.
Building on these results, data captured from a controlled post-earthquake simulation environment is used to study the feasibility of real-world implementation. High-resolution point clouds with cm-level accuracy and detailed textural information are acquired using a survey-grade laser scanner. Based on this experiment, the Mono-Temporal Disaster City 3D Dataset (C3DO) was generated. The points in the dataset are labeled as damaged or undamaged, and the area is spread across 6 buildings with an emphasis on concrete structures. Results from using C3DO demonstrated that the proposed algorithm rendered a high average MCC value of 0.79 which is measured on the scale from ���1 to +1, outperforming the current SOA algorithms. An average of 10% increase in MCC for panels in the C3DO dataset compared to GITRA.
Similar to synthetic data analysis, real-world data analysis only required knowledge about the panel thickness and had no dependence on ground truth. In a post-earthquake environment, the observed damaged region characteristics are typically unique to every occurrence, depending upon the characteristics and building design of the locality where the disaster occurs. Based on these results, it can be asserted that the proposed analysis is more reliable than the discussed state-of-the-art for the first responders or other stakeholders for damage assessment without prior knowledge of the environment
Investigations of Laser Propagation Through Realistic Turbulence Using Highly-Resolved Direct Numerical Simulations
No full textClassical theories predict the variation of distortions of electromagnetic waves as it propagates through a turbulent ���eld. Most of these theoretical developments assume planar wavefronts and idealized turbulence with statistics constant or varying smoothly along the propagation path. These assumptions, however, are rarely true in real world applications. Furthermore, there is virtually no systematic validation of these theoretical developments, especially when turbulence is variable along the propagation direction. In this work, we develop simulation codes that propagate electromagnetic waves through realistic turbulence generated from well-resolved direct numerical simulations to study the distortions in a laser beam propagating through these ���elds. Speci���cally, we look at variance of log-amplitude ���uctuations of propagating planar wavefronts and focus on two things. First, we show that the variation in statistics of instantaneous turbulent ���elds signi���cantly affects the variance of ���uctuations in propagating beams. This includes departures from theoretical scaling laws as well as oscillations not explained by existing theories. Second, we investigate the effect of sudden variation in the turbulent ���eld by essentially making the turbulent ���uctuations zero after a short propagation through turbulence. Here again we observe that the theory fails to predict the distortions in the propagating beam. Finally, we study the effects of turbulent ���ow on ���nite beams, speci���cally focused gaussian beams as it relates to the location of the centroid along the propagation distance, as well as the size and location of the beam waist. It is observed that the presence of turbulence causes the beam waist to increase in size and occur at a shorter distance when compared with vacuum propagation. Strong differences between averages and individual realizations of the propagation shed light on potential issues in predicting distortions in wavefronts using mean theories. The large high-���delity databases generated here will allow for detailed testing of existing theories and models as well as for developing new ones
Enhancing Feature Matching Performance through Monocular Depth Estimation
No full textOne key problem in computer vision is the need to find correspondences between images. For example, a major task is to identify whether two images are of the same object or view, but taken from different angles, lighting, scale, occlusion, blur, and other environmental factors. Known as the image matching problem, this is commonly solved through identifying key features of the images and matching them across different images, in a process called feature matching. With a broad range of applications in safety-critical fields such as autonomous vehicles, it is crucial that the matched pairs of features have a high degree of accuracy and that the feature matcher is robust across different challenging situations. One traditional method to identify and match features is using Scale Invariant Feature Transform (SIFT) to identify features and then applying a ratio test to find matches between the features. However, current feature matching techniques struggle in certain situations, including images of indoor, narrow spaces such as hallways. We examine ways to improve the performance of traditional feature matchers such as the SIFT + ratio test method in the hallway problem by incorporating the use of depth maps generated through monocular depth estimation. Through methods such as convolutional neural networks and Markov random fields, monocular depth estimation models can estimate the depth information of a scene from a single, 2-dimensional RGB image. By using the depth maps of a pair of images, we seek to enhance traditional feature matchers by eliminating additional matches based on a percentage difference in depth values greater than a certain threshold. In this paper, we implement the enhanced feature matcher by adding the depth filtering step after the initial match filtering process of the SIFT + ratio test method. We then examine the effectiveness of different elimination thresholds on outliers and compare the performance with that of the traditional feature matcher. We specifically compare the performance of feature matchers in addressing the hallway problem, with two images of the same hallway taken at different angles and depths
Structurally Divergent Proteins Studied by Native Ion-Mobility Mass Spectrometry: Metallothionein and Transthyretin
No full textNative mass spectrometry (MS) has become an important characterization technique in the studies of protein structures, especially for structurally divergent proteins, both intrinsically disordered proteins such as metallothioneins (MT) and proteins that are dynamics and prone to aggregation such as transthyretin (TTR). Structural information and other properties including stabilities, interactions with other proteins or ligands can be generated through MS results.
Here, we examine the effects of Ag and Cu binding on the stability and structure(s) of MT2A using chemical labelling, top-down and bottom-up proteomics as well as thermal-induced dissociation of MT-2A in solution and collision-induced unfolding of the gas phase Cu-MT complexes. The results we presented confirm that Cu6-MT preferably binds to �� domain to form a stable product, while the extra four Cu ions are bound to ��-domain less preferred but still cooperative. The mixed-metal Cd/Cu-MT and Ag/Cu-MT illustrate diversity for Cu-MT binding as well as strong cooperative binding for both metals in both domains. Lastly, results from collision-induced unfolding reveals that the stabilities of Cu-MTs increase with a higher degree of metalation. The major differences in collision-induced unfolding and thermal-induced unfolding provide more evidence for the hypothesis that [MT]^4+ and [MT]^5+ originated from very different solution phase structures. And these results raised the interest of another structurally dynamic protein TTR, that the comparison of CIU and TIU of TTR further proves the function of solvents in stabilizing protein structures. We further started the monitoring of soluble oligomers in the process of TTR aggregation of multiple physiological related mutants and validated the order of reaction rates of these mutants. But further ion mobility-based experiments should be performed to improve the identification of the high m/z species in the aggregation process
What Skills and Experiences Get Agricultural Communication Students Hired: An Industry Perspective
No full textAgricultural communications is an ever-changing field. Undergraduate students enrolled in agricultural communications programs are exposed to many different forms of communication (e.g., writing, digital media, television/radio production) and are encouraged to participate in high-impact learning experiences (e.g., internships, study abroad, research). However, many recent graduates are still ill prepared for the 21st century global workforce. This research study will surveyed agricultural communications and journalism employers who are members of the American Agricultural Editor���s Association, the ABM Agri Media Council, and the Livestock Publications Council. Participants identified key characteristics they when reviewing r��sum��s for entry-level positions in agricultural communications, how industry professionals rank soft skills and communication skills when considering hiring new employees for entry-level positions and what r��sum�� characteristics and soft skills have a higher monetary value to industry professionals. This research study found that employers seek students with a wide variety of applicable experiences and internships, strong technical and soft skill set. This study also found that communication skills and communicating accurately and concisely are ranked the highest among soft skills and communication skills and have the highest monetary value
Statewide 2022 Air Emission Calculations from Wind and Other Renewables VOL I
No full textA report to the Texas Commission on Environmental Quality for the Period January Period January 2022 ��� December 2022