Ground-penetrating radar (GPR) has applications across many domains, including archaeology, mining, and infrastructure inspection. This review is specifically focused on urban subsurface utility mapping, where accurate detection of buried pipelines, cables, and conduits is critical for excavation safety and infrastructure management. Within this scope, two major barriers are identified: event–utility mismatch and the synthetic–field domain gap. Bibliometric analysis shows increasing reliance on deep learning, yet most methods remain limited to event-level hyperbola detection rather than utility-level inference. In real urban environments, radar responses are often affected by orientation-dependent signatures, clutter, overlapping reflections, and non-utility anomalies, making detected events difficult to map directly to physical infrastructure. In parallel, models trained on synthetic data frequently show limited field generalization because simulated radargrams do not fully reproduce soil heterogeneity, acquisition variability, and system artifacts. The review argues that future progress in urban utility mapping requires a shift toward utility-level reasoning supported by multi-sensor fusion, physics-guided learning, hybrid simulation–field datasets, and uncertainty-aware interpretation. Such advances are essential for making GPR outputs more reliable and actionable in urban engineering practice.

Construction sites pose significant safety challenges due to the complex interactions between personnel and heavy machinery in dynamic environments. This article presents a novel light detection and ranging (LiDAR)-based monitoring system for tracking and safety monitoring of construction sites. The proposed system employs a comprehensive pipeline integrating point cloud processing, 3-D object detection with algorithmic background filtering, and multiobject tracking with Kalman-filter-based velocity estimation. Experimental validation demonstrates the system's capability to simultaneously track multiple objects, including construction vehicles and workers, with speed monitoring and visualization. The object detection achieves a mean average precision (mAP@0.5) of 0.733, while the tracking system maintains a multiple object tracking accuracy (MOTA) of 75.66% and multiple object tracking precision (MOTP) of 72.94%, which indicate significant potential for enhancing construction site safety through automated monitoring, providing supervisors with continuous awareness of vehicle speeds and personnel locations, thereby reducing accident risks and improving operational efficiency in construction environments. Our project is available at https://github.com/Saturn-Chao-He/Construction-Site-Tracking

Rapid and accurate building damage assessment is essential for effective post-disaster response, yet the development of reliable artificial intelligence (AI) models remains constrained by the limited availability of high-quality, operationally relevant benchmark datasets. To address this gap, this paper presents AFTERMAP (Aerial FEMA-Aligned Targeted Extraction and Reconstruction Mapping for Post-Disaster Building Damage), a benchmark UAV imagery dataset for instance segmentation of post-disaster building damage. The dataset contains 1926 high-resolution UAV images with pixel-level annotations for five building damage categories, following FEMA Preliminary Damage Assessment (PDA) guidelines where applicable: Destroyed, Major Damage, Minor Damage, Tarp, and No Damage. Using AFTERMAP, we benchmarked recent YOLO-based instance segmentation models and evaluated their performance using mAP50 and mAP50-95. Among the evaluated models, YOLO26-L achieved the best performance, obtaining a test mAP50 of 0.600 and outperforming both Mask R-CNN (0.524) and Mask2Former (0.368) in cross-architecture comparisons. Model generalization was further demonstrated through an independent case study using UAV imagery collected after the 2025 Somerset–London tornado in Kentucky, where the trained model successfully identified building-level damage patterns under real-world field conditions. The results demonstrate that AFTERMAP provides a challenging and realistic benchmark for UAV-based post-disaster building damage assessment. By combining FEMA-aligned annotations with high-resolution UAV imagery, the dataset establishes a standardized resource for developing and evaluating instance segmentation models that can support rapid post-disaster damage assessment and emergency response.

The effectiveness of first responder training relies heavily on the realism and interactivity of simulated environments. Traditional 3D animated scenes often lack the visual fidelity and detail needed to accurately represent disaster-damaged buildings, limiting trainees' ability to prepare for real-world scenarios. This paper presents a novel approach for immersive disaster response training by integrating photorealistic scene reconstruction into a game engine based virtual reality (VR) platform. Real-world damaged environments are captured and processed into high-fidelity Gaussian splat models, which are then seamlessly imported into the game engine to create highly detailed, navigable VR simulations. Trainees can explore these realistic environments, interact with contextual information hotspots, and practice navigation and assessment skills in conditions that closely mirror actual disasters. This approach has the potential to bridge the gap between virtual training and real-world preparedness, offering a flexible and cost-effective solution for enhancing disaster response readiness. Future work will focus on expanding scenario diversity and evaluating training outcomes through participant studies.

Effective post-disaster damage assessment is crucial for guiding emergency response and resource allocation. This study introduces DamageScope, an integrated deep learning framework designed to detect and classify building damage levels from post-disaster satellite imagery. The proposed system leverages a convolutional neural network trained exclusively on post-event data to segment building footprints and assign them to one of four standardized damage categories: no damage, minor damage, major damage, and destroyed. The model achieves an average F1 score of 0.598 across all damage classes on the test dataset. To support geospatial analysis, the framework extracts the coordinates of damaged structures using embedded metadata, enabling rapid and precise mapping. These results are subsequently visualized through an interactive, web-based platform that facilitates spatial exploration of damage severity. By integrating classification, geolocation, and visualization, DamageScope provides a scalable and operationally relevant tool for disaster management agencies seeking to enhance situational awareness and expedite post-disaster decision making.

In the aftermath of natural disasters, accurate assessment of structural damage and identification of critical areas are essential for effective rescue operations. This research introduces a framework that combines drone-based imaging, deep learning analytics, and augmented reality (AR) to enhance disaster response. The process begins with unmanned aerial vehicles (UAVs) capturing high-resolution images of disaster-stricken areas, including collapsed buildings and obstructed pathways. These images are processed using deep learning models specifically designed to assess damage severity and identify critical structures requiring immediate attention. The insights generated by the deep learning models are seamlessly integrated into an AR application, which overlays this information onto the physical environment. First responders equipped with mobile devices can access real-time, actionable intelligence, improving situational awareness and decision-making on the ground. Preliminary simulations and field tests demonstrate the framework's potential to streamline rescue operations and enhance disaster response capabilities. The findings demonstrate the system's potential to effectively streamline disaster management operations and enable rapid response.

Tornadoes are among the most destructive natural disasters in the United States, with over one thousand occurrences annually, causing significant human and economic losses. Rapid and accurate damage assessment is critical for effective disaster response, but traditional manual methods, relying on the Enhanced Fujita (EF) scale, are time-consuming and prone to human error. This study addresses these challenges by developing an automated tornado damage classification system using deep learning. Leveraging a curated dataset of thousands of labeled post- event images from NOAA's Storm Damage Viewer, categorized by EF0 to EF5 ratings, this work trained and evaluated deep learning models to predict damage severity. The results demonstrate the potential for scalable and accurate damage assessments, offering critical insights for emergency responders. Additionally, the study introduces a benchmark tornado damage dataset to advance future research in this domain. These contributions aim to enhance disaster response efficiency and resource allocation in tornado-affected areas.

In healthcare facilities,including hospitals,pathogen transmission can lead to infectious disease outbreaks,highlighting the need for effective disinfection protocols.Although disinfection robots offer a promising solution,their deployment is often hindered by their inability to accurately recognize human activities within these environments.Although numerous studies have addressed Human Activity Recognition(HAR),few have utilized scene graph features that capture the relationships between objects in a scene.To address this gap,our study proposes a novel hybrid multi-classifier information fusion method that combines scene graph analysis with visual feature extraction for enhanced HAR in healthcare settings.We first extract scene graphs,complete with node and edge attributes,from images and use a graph classifi-cation network with a graph attention mechanism for activity recognition.Concurrently,we employ Swin Transformer and convolutional neural network models to extract visual features from the same images.The outputs from these three models are then integrated using a hybrid information fusion approach based on Dempster-Shafer theory and a weighted majority vote.Our method is evalu-ated on a newly compiled hospital activity data set,consisting of 5,770 images across 25 activity categories.The results demonstrate an accuracy of 90.59％,a recall of 90.16％,and a precision of 90.31％,outperforming existing HAR methods and showing its potential for practical applications in healthcare environments.

Industrialized buildings, characterized by off-site manufacturing and on-site installation, offer notable improvements in efficiency, cost-effectiveness, and material use. This transition from traditional construction methods not only accelerates building processes but also enhances working efficiencies globally. Despite its widespread adoption, the performance of industrialized building manufacturing (IBM) can still be optimized, particularly in enhancing time efficiency and reducing costs. This paper explores the integration of Artificial Intelligence (AI) and robotics at IBM to improve efficiency, cost-effectiveness, and material use in off-site assembly. Through a narrative literature review, this study systematically categorizes AI-based Robots (AIRs) applications into four critical stages—Cognition, Communication, Control, and Collaboration and Coordination, and then investigates their application in the factory assembly process for industrialized buildings, which is structured into distinct stages: component preparation, sub-assembly, main assembly, finishing tasks, and quality control. Each stage, from positioning components to the integration of larger modules and subsequent quality inspection, often involves robots or human-robot collaboration to enhance precision and efficiency. By examining research from 2014 to 2024, the review highlights the significant improvements AI-based robots have introduced to the construction sector, identifies existing challenges, and outlines future research directions. This comprehensive analysis aims to establish more efficient, precise, and tailored construction processes, paving the way for advanced IBM.

Ground-penetrating radar (GPR) is a widely utilized technique for subsurface imaging and object detection. Despite its broad applications, GPR data preparation and analysis present challenges, particularly when realistic test environments are unavailable, or when simulation data lacks accuracy. This research addresses these challenges by developing a controlled test bed at the KSU Field Station, incorporating soil layers of varying densities and embedded objects positioned at known locations. The test bed is composed of river sand and fine-grained Piedmont soil, with flexible configurations allowing for the inclusion of buried objects such as pipes of various sizes made from PVC and metal, simulating realistic subsurface environments. Real GPR data was collected from multiple layers within this constructed underground infrastructure. In addition, synthetic data was generated using gprMax simulations to explore signal characteristics. The research highlights the discrepancies between real and simulated GPR data, emphasizing their significance for improving subsurface mapping and object detection. The experimental data generated from this study will provide valuable insights for geotechnical engineers, construction professionals, and field operators, helping to prevent issues such as undetected voids or utility line strikes. By addressing limitations in GPR data preparation and analysis, this research enhances the accuracy and reliability of subsurface imaging, with potential applications across industries such as civil engineering and construction.

Ground penetrating radar (GPR) is an advanced nondestructive testing (NDT) technique extensively used for assessing the structural integrity of concrete bridge decks. Despite its effectiveness, the manual interpretation required for GPR data hampers its broader application in bridge deck evaluations. This study presents a novel automated workflow that utilizes GPR scans to produce detailed deterioration maps of bridge decks, significantly enhancing the efficiency and precision of inspections. The automated process involves detecting rebar regions within the GPR data based on their unique electromagnetic signatures and employing hyperbola clustering to accurately localize each rebar by identifying the peaks of hyperbolic patterns. This crucial step aids in assessing the structural integrity of the bridge. The automated workflow culminates with the creation of deterioration maps that highlight potential structural weaknesses, enabling targeted maintenance and repairs. Field experiments conducted on a bridge deck confirmed the method's effectiveness, showcasing its potential to expedite the bridge inspection process significantly. The development of this automated processing pipeline marks a substantial advancement in the application of GPR technology in civil engineering, paving the way for more reliable and streamlined bridge maintenance practices.

Housing recovery represents a critical component of disaster recovery, and accurately forecasting household relocation decisions is essential for guiding effective post-disaster reconstruction policies. This study explores the use of machine learning algorithms to improve the prediction of household relocation in the aftermath of disasters. Leveraging data from 1304 completed interviews conducted as part of the Displaced New Orleans Residents Survey (DNORS) following Hurricane Katrina, we evaluate the performance of Logistic Regression (LR), Random Forest (RF), and Weighted Support Vector Machine (WSVM) models. Results indicate that WSVM significantly outperforms LR and RF, particularly in identifying the minority class of relocated households, achieving the highest F1 score. Key predictors of relocation include homeownership, extent of housing damage, and race. By integrating variable importance rankings and partial dependence plots, the study also enhances interpretability of machine learning outputs. These findings underscore the value of advanced predictive models in disaster recovery planning, particularly in geographically vulnerable regions like New Orleans where accurate relocation forecasting can guide more effective policy interventions.

Effective and timely disaster response is critical for minimizing damage and saving lives following catastrophic events. This paper presents a novel framework that integrates immersive simulations, synthetic data generation, and AI-enhanced robotic training to support post-disaster operations. The framework comprises three main components. First, immersive virtual environments are developed to train first responders by simulating complex disaster scenarios, aiming to enhance decision-making and reduce response time. Second, synthetic datasets derived from these simulations are used to train artificial intelligence models for damage assessment and resource allocation. Third, Unity-based environments are employed to train autonomous robots to navigate debris-filled areas, facilitating improved human- robot collaboration during search and rescue missions. Together, these components form a scalable and cost-effective solution to strengthen emergency preparedness and operational efficiency in high-risk environments.

Automated structural defect detection is essential for ensuring the safety and maintenance of civil infrastructure, particularly in bridges where defects such as cracks, spalling, and corrosion can compromise structural integrity. This paper presents a comparative study of three semantic segmentation models—U-Net, Feature Pyramid Network (FPN), and DeepLabv3+—for detecting and classifying structural defects in bridge imagery. Each model was evaluated using two encoder architectures, EfficientNet B3 and MobileOne S4, to assess the impact of different feature extraction strategies on segmentation accuracy. The experiments were conducted using the DACL benchmark dataset, which includes a wide range of defect classes. FPN paired with EfficientNet B3 demonstrated the highest mean accuracy across most defect categories, outperforming the other combinations, particularly in detecting common defects such as cracks and graffiti. However, certain defect types, such as hollow areas and cavities, presented challenges for all models. These results highlight the effectiveness of deep learning models in automated defect detection, while also identifying areas where further refinement is needed to improve performance in more complex defect scenarios.

Bridge infrastructure in the United States is aging, necessitating efficient and accurate inspection methods. Ground-penetrating radar (GPR) is a widely used non-destructive testing (NDT) method for detecting subsurface anomalies in bridge decks. However, manual interpretation of GPR scans is labor-intensive, and annotated datasets for deep learning applications are limited. This study investigates YOLO-based deep learning models for automated rebar detection using a combination of real and synthetic GPR data. A dataset comprising 2255 real GPR images from four bridges and 20,000 simulated GPR scans was used to train and evaluate YOLOv8, YOLOv9, YOLOv10, and YOLOv11 under different training strategies. The results show that pretraining on simulated GPR data improves detection accuracy compared to conventional COCO pretraining, demonstrating the effectiveness of domain-specific transfer learning. These findings highlight the potential of simulated GPR data for training deep learning models, reducing reliance on extensive real-world annotations. This study contributes to AI-driven infrastructure monitoring, supporting the development of more scalable and automated GPR-based bridge inspections.

Natural disasters often magnify pre-existing socioeconomic disparities, underscoring the critical need to examine the underlying factors that contribute to unequal impacts across communities. This study analyzes the socioeconomic and demographic determinants of building damage severity during Hurricane Katrina in New Orleans. Drawing on data from the Displaced New Orleans Residents Survey (DNORS), a multinomial logistic regression model is employed to estimate the likelihood of households experiencing varying levels of structural damage. The analysis identifies race, homeownership status, insurance coverage, employment status, household composition, and the presence of children as significant predictors of damage severity. These findings highlight the complex interplay between social vulnerability and disaster outcomes, revealing how structural inequalities shape exposure and recovery capacity. The study emphasizes the importance of equity-focused policy interventions that strengthen housing resilience, expand access to financial protections, and promote inclusive disaster preparedness. By incorporating these insights, policymakers can better support at-risk populations and advance community resilience in the face of future hazards.

Social media has become an indispensable resource in disaster response, providing real-time crowdsourced data on public experiences, needs, and conditions during crises. This user-generated content enables government agencies and emergency responders to identify emerging threats, prioritize resource allocation, and optimize relief operations through data-driven insights. We present an AI-powered framework that combines natural language processing with geospatial visualization to analyze disaster-related social media content. Our solution features a text analysis model that achieved an 81.4% F1 score in classifying Twitter/X posts, integrated with an interactive web platform that maps emotional trends and crisis situations across geographic regions. The system’s dynamic visualization capabilities allow authorities to monitor situational developments through an interactive map, facilitating targeted response coordination. The experimental results show the model’s effectiveness in extracting actionable intelligence from Twitter/X posts during natural disasters.

This study pioneers the use of synthetic image datasets, generated via the Unity game engine, to train deep learning models for construction work zone scene understanding. This innovative approach simplifies data acquisition and ensures a rich, diverse dataset that includes various construction scenarios, both hazardous and typical. We created 12,360 images with accurate bounding box annotations, ensuring high-quality, consistent data crucial for model training and validation, and effectively addressing the ambiguities often found in real-world datasets. An object detection benchmark was established using this dataset alongside eight state-of-the-art object detectors. This benchmark thoroughly evaluates the performance of these detectors on a wide range of construction site images, enabling comparisons and analyses of different models. It highlights their respective strengths and weaknesses in construction site applications. Notably, YOLOv8-L demonstrated exceptional performance, achieving a mean average precision (mAP) of 70.7% on the validation set and 69.8% on the testing set. These results underscore the efficacy of synthetic datasets in training models for complex scene understanding. This integration of synthetic and real-world imagery has the potential to revolutionize scene comprehension in construction zones, significantly enhancing safety and efficiency in the construction industry.

Bridge inspection is a crucial process for ensuring the safety and reliability of transportation infrastructure. Traditional bridge inspections are time-consuming, costly, and often require bridges to be closed, disrupting traffic. In recent years, the use of drones and computer vision techniques for bridge inspection has gained attention due to their ability to provide accurate and comprehensive data while reducing costs and disruptions. This paper presents an automated bridge inspection framework that utilizes drones and computer vision techniques for detecting and analyzing cracks on bridge decks. The framework comprises three main components: orthomosaic map generation, deep learning-based crack detection, and georeferencing and visualization in a geographic information system (GIS) platform. The cracks are segmented, identified, and extracted with their georeferenced coordinates, which can be seamlessly integrated into a GIS platform. This integration enables enhanced visualization and spatial analysis of the cracks. In addition, an image data set has been created to facilitate the process of crack segmentation in the context of the proposed automated bridge inspection framework. The network achieved a mIoU of 80.5%, a dice coefficient of 88.1%, a precision of 77.5%, and a recall of 76.5%, highlighting the robust performance of the network in crack detection. The proposed framework was evaluated on a real bridge, and the results showed that it detected and analyzed cracks accurately and efficiently. This framework can be adaptable to various types of infrastructure, making it a valuable tool for managing transportation infrastructure.