<rss version="2.0" xmlns:atom="https://www.w3.org/2005/Atom">
  <channel>
    <title>Transport Research International Documentation (TRID)</title>
    <link>https://trid.trb.org/</link>
    <atom:link href="https://trid.trb.org/Record/RSS?s=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" rel="self" type="application/rss+xml" />
    <description></description>
    <language>en-us</language>
    <copyright>Copyright © 2026. National Academy of Sciences. All rights reserved.</copyright>
    <docs>http://blogs.law.harvard.edu/tech/rss</docs>
    <managingEditor>tris-trb@nas.edu (Bill McLeod)</managingEditor>
    <webMaster>tris-trb@nas.edu (Bill McLeod)</webMaster>
    <image>
      <title>Transport Research International Documentation (TRID)</title>
      <url>https://trid.trb.org/Images/PageHeader-wTitle.jpg</url>
      <link>https://trid.trb.org/</link>
    </image>
    <item>
      <title>Autonomous Robotic System for Detecting Rail Anchor Failures Using Computer Vision</title>
      <link>https://trid.trb.org/View/2772533</link>
      <description><![CDATA[This project addresses the issue of rail anchor failures that pose significant risks to rail safety. The anchor failure can cause track misalignment, rail buckling, and derailments and adds to the maintenance costs. The proposed innovation is an autonomous robotic system designed to detect and locate rail anchor failures with precision and efficiency. By leveraging computer vision and machine learning, the system will identify problems such as slippage, damage, or displacement of rail anchors in real-time and provide precise location mapping for targeted maintenance interventions. The research approach will involve creating a high-resolution annotated image dataset and training a deployable artificial intelligence (AI) model to enable accurate, consistent, and scalable anchor inspections — capabilities that are not adequately addressed by existing systems. The work will be performed in two phases. The first phase will complete the design and development of the robotic platform equipped with high-resolution cameras and AI-based diagnostic tools. In the second phase, experimental validation of the system will be carried out on a railway track to ensure scalability, reliability, and effectiveness under real-world conditions. The system will automate the inspection process, eliminate human error, and enable proactive maintenance resulting in reduced downtime and enhanced rail safety. The scalable design and integration of advanced diagnostics will make the system suitable for adoption by the railroad industry on a wider scale and contribute to enhanced rail infrastructure reliability and operational efficiency.]]></description>
      <pubDate>Fri, 04 Sep 2026 08:28:41 GMT</pubDate>
      <guid>https://trid.trb.org/View/2772533</guid>
    </item>
    <item>
      <title>Research on Intelligent Recognition of Subgrade Compactness Based on Deep Learning</title>
      <link>https://trid.trb.org/View/2712011</link>
      <description><![CDATA[Subgrade compaction is a critical metric for assessing subgrade quality. The traditional compaction detection method is complicated and time-consuming. Based on the deep learning method, this study proposes an image-based compaction measurement (IBCM) method, constructs four improved convolutional neural network (CNN) models, including the improved VGG16, ResNet50, DenseNet121, and MobileNetV1 models, and applies these models to the sandy loam compacted soil sample images under different compaction conditions to establish the intelligent recognition method of subgrade compaction. Through a comprehensive comparison of the performance of these models in the task of intelligent identification of compactness, the improved VGG16 model with the best performance is finally selected, and its test set accuracy can reach 0.9733. Simultaneously, laboratory and field tests validate the reliability and practicality of this approach. The accuracy rates achieved in these tests are 0.9800 for laboratory assessments and 0.9400 for field evaluations. Consequently, this advanced method of recognizing compaction levels through deep learning can be effectively utilized in real-world engineering applications. This not only enhances the efficiency of compaction degree evaluations but also significantly contributes to the enhancement of subgrade quality.]]></description>
      <pubDate>Fri, 28 Aug 2026 13:34:20 GMT</pubDate>
      <guid>https://trid.trb.org/View/2712011</guid>
    </item>
    <item>
      <title>Robustness of Traffic Anomaly Detection in UAV Videos: Cross–Weather Generalization</title>
      <link>https://trid.trb.org/View/2698357</link>
      <description><![CDATA[Recent advances in deep learning have enabled accurate video anomaly detection for traffic monitoring and aerial surveillance. However, robust detection in UAV videos captured under adverse weather remains challenging due to severe visual degradations and domain shifts. This paper presents a systematic robustness study of six representative anomaly detection methods—Future Frame Prediction, Spatio-Temporal Dissociation, MNAD, MLEP, ANDT, and ASTT—on two UAV traffic datasets (UIT-ADrone and Drone-Anomaly). To evaluate cross-weather generalization in a controlled manner, we construct adverse-weather variants using established image-to-image translation models for fog, rain, and snow. We report frame-level performance using ROC-AUC and Equal Error Rate (EER) under (i) cross-weather testing and (ii) cross-dataset transfer settings. Results consistently show notable degradation across all models under adverse weather; CNN-based approaches tend to be more resilient than Transformer-based ones under heavy visibility loss and noise patterns. Our findings highlight failure modes that commonly arise in adverse weather and provide practical insights for designing more robust UAV-based traffic anomaly detection systems.]]></description>
      <pubDate>Fri, 28 Aug 2026 08:35:07 GMT</pubDate>
      <guid>https://trid.trb.org/View/2698357</guid>
    </item>
    <item>
      <title>A kernelized deep regression method to simultaneously predict and normalize displacement responses of long-span bridges via limited synthetic aperture radar images</title>
      <link>https://trid.trb.org/View/2695841</link>
      <description><![CDATA[Synthetic aperture radar (SAR) images retrieved by spaceborne remote sensing have recently gained significant attention as an affordable and effective solution to provide structural responses in terms of displacements from field measurements. Notwithstanding, this process may lead to partial/scattered information due to the limitations of SAR images. Furthermore, the effects of unmeasured environmental and/or operational conditions on structural responses and sensitivity of SAR-extracted displacements of full-scale structures like long-span bridges to these conditions still stand as major challenges. In this work, an innovative machine learning-aided methodology is put forward for handling these issues. The proposed methodology simultaneously predicts and normalizes displacement data within a two-stage kernelized deep regression (KDR) framework. The first stage involves kernelized regressor modeling and selection, exploiting Gaussian process regression and support vector regression. The second stage is based on deep regressor modeling via a long-short-term-memory neural network. The proposed methodology is shown to display high accuracy in prediction limited displacement data independent of unmeasured environmental/operational data. To concretely assess the performance of the proposed methodology, displacement responses from two long-span bridges and seasonal temperature records are considered. Results show that the approach is superior to available state-of-the-art techniques.]]></description>
      <pubDate>Fri, 28 Aug 2026 08:35:05 GMT</pubDate>
      <guid>https://trid.trb.org/View/2695841</guid>
    </item>
    <item>
      <title>Boundary-Guided Real-Time Semantic Segmentation and Pixel-Level Quantification of Pavement Cracks</title>
      <link>https://trid.trb.org/View/2685891</link>
      <description><![CDATA[Timely and accurately extracting and assessing pavement cracks is crucial for intelligent transportation systems (ITS) to improve road maintenance and safety. In this paper, we present an automated framework for crack semantic segmentation and quantification using optical images. First, a unique boundary-guided real-time high-resolution network is proposed, termed as BulletNet, for crack semantic segmentation. BulletNet is a bullet-head structure that can retain crack details while ensuring real-time inference speed, in which a Cross-Scale Global Attention (CSGA) module is designed to enhance global feature representation and pixel-level relations, as well as a Boundary-Guided Fusion (BGF) module proposed to utilize boundary features to guide the fusion of crack details and contextual information. Second, a Pixel-level Crack Quantification (PCQ) algorithm is proposed for complex cracks, incorporating an Improved Discrete Skeleton Evolution (IDSE) method to optimize skeleton pruning for accurate crack length and a normal vector correction method to adjust propagation direction for precise crack width. Comprehensive experiments on three datasets showed that the proposed BulletNet surpassed the comparative models in terms of efficiency and performance, with average F1-score, mIoU, and Frames per second (FPS) of 87.20%, 88.70%, and 125.53, respectively. In addition, tested on 200 images, the PCQ calculated the crack maximum widths and lengths with an average relative error of 6.96% and 4.62%, respectively. Finally, BulletNet was deployed on edge devices for field testing, and a system based on the PCQ algorithm was developed to validate the effectiveness of the entire framework.]]></description>
      <pubDate>Thu, 27 Aug 2026 16:33:40 GMT</pubDate>
      <guid>https://trid.trb.org/View/2685891</guid>
    </item>
    <item>
      <title>Using Machine Learning with Photolog Images to Identify Opportunities to Improve Highway Safety</title>
      <link>https://trid.trb.org/View/2768417</link>
      <description><![CDATA[Roadway features captured in Kentucky Transportation Cabinet (KYTC) Photolog imagery are relevant to safety, operations, design, construction, maintenance, and system management for a large portion of Kentucky’s state-maintained roadway network. However, identifying conditions of interest hinges largely on manual review. Given the immense volume of imagery and competing demands placed on staff, reliance on manual review limits the Cabinet’s ability to use Photolog imagery in a scalable, consistent, and timely manner. As a result, roadway conditions that affect safety, operations, and asset management may go unidentified or are located only after a delay. Not having efficient methods to screen Photolog imagery systematically prevents KYTC from fully leveraging this resource to make data-informed decisions about transportation system management.]]></description>
      <pubDate>Wed, 26 Aug 2026 17:04:24 GMT</pubDate>
      <guid>https://trid.trb.org/View/2768417</guid>
    </item>
    <item>
      <title>Drilled Shaft Imaging With 2D Ultrasonic Waveform Tomography</title>
      <link>https://trid.trb.org/View/2761028</link>
      <description><![CDATA[Drilled shafts are widely used as deep foundations for transportation infrastructure; however, construction defects such as soil intrusion, concrete segregation, slurry inclusions, necking, and soft-bottom conditions can compromise structural integrity and long-term performance. Conventional crosshole sonic logging (CSL) provides valuable quality-control information but offers limited capability for imaging shaft geometry and defect extent. This study developed, optimized, and field-validated a two-dimensional ultrasonic full waveform Inversion (UFWI) framework for high-resolution drilled shaft imaging using CSL waveform data. The method combines acoustic wavefield modeling, adjoint-state inversion, and regularized model updating to reconstruct cross-sectional P-wave velocity distributions without prior knowledge of shaft boundaries or defect locations. Synthetic studies demonstrated that UFWI can accurately image shaft geometry and detect internal and boundary defects. Parametric analyses further showed that at least five CSL access tubes are required for reliable imaging, while approximately one tube per foot of shaft diameter is recommended for shafts 6 ft in diameter or larger. Field validation was conducted on a full-scale 6-ft-diameter, 40-ft-long drilled shaft constructed at the Florida Department of Transportation (FDOT) Hawthorne test facility with intentionally embedded defects, including concrete segregation, necking with soil intrusion, slurry inclusions, and a soft-bottom condition. UFWI successfully reconstructed shaft boundaries; identified major defect horizons; and localized anomalies consistent with documented as-built conditions. Compared with conventional CSL, UFWI provided higher spatial resolution and more detailed characterization of defect geometry, demonstrating its potential as a quantitative nondestructive evaluation tool for drilled shaft quality assurance and integrity assessment.]]></description>
      <pubDate>Wed, 26 Aug 2026 09:22:37 GMT</pubDate>
      <guid>https://trid.trb.org/View/2761028</guid>
    </item>
    <item>
      <title>Fundamental Study on Crack Detection Method for Prestressed Concrete Sleepers Using Deep Learning Model</title>
      <link>https://trid.trb.org/View/2709145</link>
      <description><![CDATA[Prestressed concrete sleepers are an important component of railway tracks, contributing to the speed and safety of train operations. Cracks appearing in the longitudinal direction of some prestressed concrete sleepers in recent years due to alkali-silica reactions have raised concerns about the efficiency of their maintenance. Therefore, this study proposes the use of a deep learning model to estimate the position and length of cracks on top surface images of prestressed concrete sleepers, as captured by a camera mounted on a maintenance vehicle. The applicability test confirmed that the method can accurately estimate the position and length of cracks in prestressed concrete sleepers, while minimizing the likelihood of false detection of ballast and fastening devices. In addition, it was demonstrated that this method can be employed to identify areas with a high concentration of cracks and analyze crack patterns on commercial lines.]]></description>
      <pubDate>Tue, 25 Aug 2026 09:54:07 GMT</pubDate>
      <guid>https://trid.trb.org/View/2709145</guid>
    </item>
    <item>
      <title>Development of a Method for Diagnosing Deterioration of Wayside Equipment using Forward-facing Train Images</title>
      <link>https://trid.trb.org/View/2709144</link>
      <description><![CDATA[Wayside equipment is installed both within stations and at different locations between them, making inspection and management labor-intensive. With the decline in maintenance staff, improving efficiency in equipment management has become essential. To address this issue, we developed a system that enables remote monitoring of trackside equipment using forward-facing train images captured by an onboard camera. The system estimates kilometrage, automatically detects equipment, and estimates deterioration of signalling equipment boxes. This paper presents the system overview, accuracy evaluation, and prospects for long-term deterioration monitoring.]]></description>
      <pubDate>Mon, 24 Aug 2026 14:55:44 GMT</pubDate>
      <guid>https://trid.trb.org/View/2709144</guid>
    </item>
    <item>
      <title>Engineering-oriented automated segmentation and quantitative analysis of building structural cracks: A multi-mechanism optimized YOLOv11-seg approach</title>
      <link>https://trid.trb.org/View/2752193</link>
      <description><![CDATA[During the long-term service of existing buildings, the initiation and propagation of structural cracks are critical indicators for evaluating building safety and durability. Computer vision techniques have achieved substantial progress in crack detection for bridges and road pavements, and considerable research has also been carried out on building crack identification. However, due to the complex morphology, diverse surface textures, and complex background environments of building cracks, most existing models trained for traffic infrastructure cannot be directly well adapted to practical building scenarios. In addition, existing public crack datasets are mostly oriented to road and bridge engineering, lacking targeted samples and refined feature descriptions for building structural cracks, which makes it difficult to support high-precision segmentation and geometric parameter quantification in building engineering. To address these application gaps, this study firstly constructs a dedicated building crack segmentation dataset BCrack containing 450 multi-scene crack images. On this basis, an improved YOLOv11-seg model (YOLO-MDAC) with multi-mechanism optimization is proposed. The mAP50 is increased from 80.2% to 83.4%, effectively improving the accuracy of crack edge segmentation. Furthermore, the proposed method is integrated with a quantitative calculation module for crack length and width. Field test results show that the measurement errors can be controlled within 5%. This research provides an efficient and reliable technical means for structural damage assessment of existing buildings, and presents good engineering application and promotion value.]]></description>
      <pubDate>Mon, 24 Aug 2026 08:44:55 GMT</pubDate>
      <guid>https://trid.trb.org/View/2752193</guid>
    </item>
    <item>
      <title>Coupling digital image correlation with dynamics metrics for fracture assessment in lightly reinforced concrete</title>
      <link>https://trid.trb.org/View/2752101</link>
      <description><![CDATA[Reliable assessment of damage in lightly reinforced concrete structures requires monitoring approaches that capture both local fracture behaviour and its impact on global structural performance (e.g., natural frequencies). Conventional vibration-based Structural Health Monitoring (SHM) methods are effective for detecting stiffness degradation but offer limited insight into crack mechanics. In contrast, vision-based techniques such as Digital Image Correlation (DIC) provide high-resolution measurements of crack initiation and propagation, yet do not directly quantify the associated loss in structural capacity. This study presents an integrated monitoring framework that combines full-field DIC with vibration-based frequency measurements to investigate their relationship. Local crack initiation, propagation, and crack opening displacement are quantified using DIC and linked to global stiffness degradation through shifts in natural frequency obtained from dynamic testing via hammer impact excitation. An integrated fracture-based model is further employed to interpret how observed crack geometry influences structural stiffness. Results demonstrate that increasing crack opening leads to measurable reductions in natural frequency, with the most pronounced changes occurring during the changeable crack rotation phase prior to failure. By experimentally linking local fracture processes to global modal behaviour, this study provides a stronger physical basis for interpreting dynamic response data, particularly in vehicle-assisted SHM applications. The proposed approach enables more reliable, mechanism-informed drive-by damage detection in reinforced concrete structures.]]></description>
      <pubDate>Mon, 24 Aug 2026 08:44:55 GMT</pubDate>
      <guid>https://trid.trb.org/View/2752101</guid>
    </item>
    <item>
      <title>CREATE Year1 Dataset for Combining Structural and Cathodic Protection with Titanium Alloy Bars [supporting dataset]</title>
      <link>https://trid.trb.org/View/2742394</link>
      <description><![CDATA[This data archive is for the Year 1 CREATE Project with support from BIL UTC under Grant No. 69A3552348330. The archive centralizes salient experimental measurements, metadata, and outputs generated by the research team. It includes: (1) Electrochemical data:  high-resolution images of specimens and test conditions, polarization, bulk resistivity, and open-circuit half-cell potential measurements from multiple specimen types. (2) Structural data: high-resolution images of specimens, material test results, pullout specimen force-deformation responses, beam specimen strain measurements and load-deformation responses. (3) Supporting documentation: detailed scaled drawings of specimens and instrument configurations, data dictionaries, and README files for reproducibility. Data are organized in a standardized directory structure with descriptive naming conventions, and file formats suitable for long-term accessibility, reuse, and future modeling efforts. All datasets are backed up securely and accompanied by descriptive metadata to ensure transparency, traceability, and compliance with CREATE archive standards.]]></description>
      <pubDate>Mon, 24 Aug 2026 08:44:55 GMT</pubDate>
      <guid>https://trid.trb.org/View/2742394</guid>
    </item>
    <item>
      <title>Surface fatigue crack segmentation in steel box girder bridges using an improved Unet convolutional neural network</title>
      <link>https://trid.trb.org/View/2742418</link>
      <description><![CDATA[Despite the success of crack segmentation, the traditional segmentation strategy and networks are not necessarily suitable for fatigue cracks with unique visual features. Towards the indistinct crack features, a network named Fatigue Crack Unet (FC-Unet) was proposed, where the novel split attention mechanism was incorporated to boost diverse feature extraction. Scaled by various ratios, raw images were cropped and constituted three datasets to represent different segmentation strategies. FC-Unet was trained and tested on each dataset, where the class imbalance problem was alleviated by the adjustment of loss functions. Results showed that the segmentation performance is mainly limited to noise interference. Preserving abundant global contexts to distinguish noises, the global inference is the superior strategy to the previous local inference. Leveraging the wide receptive field in global inference, the down-sampling rate can be decreased to save resolution losses. Combining the focus-based and region-based loss functions, the accuracy of imbalanced data was further improved. Compared with nine classical networks, the proposed FC-Unet enjoyed a powerful backbone and achieved a superior metric of 83.1% mean intersection over union (MIoU).]]></description>
      <pubDate>Mon, 24 Aug 2026 08:44:54 GMT</pubDate>
      <guid>https://trid.trb.org/View/2742418</guid>
    </item>
    <item>
      <title>Aerial Intelligence: Road Safety Evaluation through AI and Machine Learning Using Drone Imagery</title>
      <link>https://trid.trb.org/View/2729652</link>
      <description><![CDATA[This paper implements the CNN-LSTM hybrid network to address the limitations of the earlier deep learning strategies such as YOLOv3, Faster R-CNN, and a single LSTM and enhance the analysis and classification of the spatiotemporal data. Classical models are competent at one or two things, such as the ability to detect spatial characteristics or the ability to identify sequential patterns. But they are having issues in establishing the appropriate balance between speed, accuracy, and performance in the real-time. The suggested CNN-LSTM hybrid improves performance on a variety of datasets by combining convolutional layers for extracting spatial details with LSTM units for modeling temporal sequences. The experimental test demonstrates that the proposed model is more accurate (95.79%) than YOLOv3 (89.0%), Faster R-CNN (91.2%), and LSTM-based sequence analysis (90.5%). The model is adaptable and precise, rendering it applicable in complex real-world scenarios such as intelligent decision-making systems, healthcare monitoring, and traffic surveillance. The results of this study show that combining spatial and temporal deep learning methods can greatly increase the accuracy and reliability of categorization. This is one method that smart systems could be built in the future.]]></description>
      <pubDate>Fri, 21 Aug 2026 17:00:58 GMT</pubDate>
      <guid>https://trid.trb.org/View/2729652</guid>
    </item>
    <item>
      <title>Examination of image conditions using multiway analysis of variance for predicting high-crash-risk intersections with image recognition AI</title>
      <link>https://trid.trb.org/View/2742883</link>
      <description><![CDATA[Traffic crashes on residential roads in Japan have declined only marginally in recent years, indicating a stagnation in safety improvements. A key challenge is the early identification of high‑risk intersections, yet existing image‑based prediction studies have not statistically validated whether specific image‑capture conditions influence model performance. This study addresses this gap by developing a convolutional neural network (CNN) model trained on on‑site photographs collected in Fukuoka City, which serves as the empirical case study. The model’s performance is evaluated using the mean crash risk score, and the effects of three image‑capture factors such as resolution, sky editing, and camera distance, are statistically examined using multiway analysis of variance (ANOVA). Results show that the model achieved a mean crash risk score of 0.776 under the tested image conditions, indicating that the Artificial Intelligence (AI) system classified high‑risk intersections with an average confidence of 77.6%. Statistical testing revealed that sky editing and camera distance significantly affected prediction reliability, with the most favorable conditions within this dataset observed at 1,280 × 960‑pixel resolution, unedited sky color, and a 10‑m camera distance. These findings provide empirical evidence that specific image‑capture settings materially influence AI‑based intersection risk prediction. Rather than proposing a superior model architecture, this study contributes a statistically validated framework for assessing how image‑capture conditions affect model outputs and offers practical guidance for standardizing image collection in transportation planning and road‑safety diagnostics.]]></description>
      <pubDate>Fri, 21 Aug 2026 08:37:10 GMT</pubDate>
      <guid>https://trid.trb.org/View/2742883</guid>
    </item>
  </channel>
</rss>