Image captioning for automated bridge inspection: a feasibility study

With the increasing expansion and improvement of the transportation network, the durability of bridges, as an important transportation infrastructure connecting urban and rural areas and crossing rivers and canyons, is directly related to the safety of people's lives and properties and the stable development of society. This paper firstly introduces the background and significance of bridge inspection technology in detail, and then deeply analyzes the current problems and challenges in bridge inspection and evaluation. Then, this paper elaborates the research contents and methods of this paper, including the comprehensive and accurate inspection of bridges using advanced inspection technology, the processing and analysis of inspection data using big data and artificial intelligence technology, and the construction of bridge safety assessment model. Through these studies, this paper proposes an efficient bridge inspection and assessment method to provide a strong guarantee for the safe operation of bridges. Through experimental investigation, the accuracy of its bridge inspection technology fluctuates between 93.5% and 98.4%; while the efficiency is always above 80%, which fully demonstrates the high efficiency of ultrasonic technology in bridge inspection. This paper finds that the efficiency and accuracy of bridge inspection can be greatly improved by adopting advanced inspection technology. At the same time, combined with big data and artificial intelligence technology, rapid processing and analysis of bridge inspection data can be realized, providing strong support for bridge safety assessment.

As the number of bridges grows in China, bridge inspection is very necessary to ensure public transport safety. With the development of various technologies in recent years, such as UAV (Unmanned Aerial Vehicle), computer vision, advanced sensing, artificial intelligence and so on, intelligent technologies in bridge inspection have developed rapidly and are gradually replacing traditional methods. Here we proposed parallel systems for bridge inspection, which introduces parallel theory into the field of bridge inspection to solve the problems of dataset shortage and special scene prediction. Based on the classification dataset (CCD) and parallel classification dataset (PCD), ConvNeXt and other networks are trained and compared. The final crack identification accuracy reached 99.22%. We believe that the framework proposed in this paper can improve the efficiency and accuracy of bridge inspection significantly.

Advances in Machine Learning (ML) and the Internet of Things (IoT) have led to the development of self-sufficient systems to check the infrastructures and maintenance processes used to maintain the vital infrastructures in the traditional monitoring and maintenance procedures. This chapter discusses how the ML algorithms and IoT sensors may be integrated and used to improve the accuracy, efficiency, and safety of the infrastructure inspection. Implementation of autonomous systems, e.g., drones and robots, enables collecting real-time data, functioning permanently, and making predictions related to maintenance reducing maintenance labour, operational cost, and downtime. A number of case studies are presented to demonstrate practical use of such technologies, such as bridge inspection through drones, predictive maintenance of railways, and intelligent pipeline monitoring. The chapter further lists the challenges fixed to such systems including; sensor calibration, data integration and data processing in real time.

Translation is a matter of absolute necessity in the globally united and yet; linguistically and culturally separated world in which we live. Translation allows individuals who speak different languages to communicate and understand one another. The obvious limitations in human translation in terms of speed, scalability and high cost gave birth to the quest for alternative means of translation which is machine translation. Machine translation involves the use of computers to translate from one natural language (source) to another natural language (target). Machine translation applications can be developed using rule-based, example- based and statistical-based technologies. Machine translation applications have been developed to translate English language text to Igala language text using rule-based and example-based approaches. Recently, Neural Machine Translation (NMT) emerged as a new paradigm that swiftly superseded other methods of machine translation evolved with the development of deep learning. This study is aimed at developing a system that can translate contents rendered in English language to Igala language using the Neural Machine Translation with attention mechanism approach. The Neural Machine Translation model for English-to-Igala translation was built with encoder-decoder architecture with an attention layer. The model was trained with a dataset using English-Igala parallel corpus that contains 50,000 parallel sentences. The dataset was partitioned into training, validation and test set in the ratio of 80/20 percent. The model was implemented using python programming language. After training, validation and testing, the output of the model was tested on a corpus of 276 selected English texts using the Bilingual Evaluation Understudy (BLEU) method for evaluating Machine Translation system. An accuracy of 71.0% was obtained. The adoption of this model will play crucial role in facilitating everyday interaction, information exchange and active participation in society among Igala people. It will give Igala people access to information in Igala language that will enhance their standard of living as it will enable them to participate in abundant business opportunities available in the online community, and ensure proper integration in the emerging information society.

Unmanned aerial vehicles (UAVs) have the potential to reduce bridge inspection time and cost while increasing safety. However, UAV-collected field data has inherent properties that complicate damage assessment. In this article, the authors integrate UAV-collected imagery data with automatic defect detection to create a novel first-pass bridge inspection algorithm, which aims to conduct an initial corrosion assessment to determine if further inspection is needed. The authors use UAV-captured images of bridges near Atlanta, Georgia, USA, to create a dataset representative of bridge inspections, including the presence of chaos and misleading objects. The proposed methodology integrates deep learning methods (fully convolutional network (FCN)) to remove natural elements in the image background that resemble corrosion, image processing techniques to quantify texture and reduce lighting effects, and unsupervised learning (K-means) for corrosion segmentation. Experimental results show that the K-means algorithm outperforms other segmentation methods, including image thresholding and deep learning, with a recall of 0.78 and mIoU of 0.72 on UAV-collected field data. Thus, the newly developed method is a promising tool to improve the efficiency and safety of bridge inspections by reducing the number of full inspections conducted on structurally sound bridges.

Bridge management systems (BMSs) are widely used to assist an inspector in performing element-level bridge inspection. Retrieving and determining target elements to be inspected becomes an important factor in the efficiency of bridge inspection. This paper presents an enhanced information retrieval (IR) method based on ontology to predict the target elements. The novelty of this method is that an improved seven-step method based on automatic mapping technology is proposed to construct a new bridge inspection ontology (BIontology), which provides a knowledge base for the present IR method. A further novelty is that a new software architecture is designed for integrating ontology, and a promising prototype system based on the software architecture is developed to realize the present IR method using SPARQL query. In addition, a novel prediction algorithm based on the present IR method is proposed to automatically recommend the target elements. A case study of ontology construction is performed to demonstrate that the improved seven-step method can accelerate the construction of the BIontology compared with the manual method. A case study of bridge inspection is implemented to verify that the proposed algorithm outperforms an existing method, thereby validating the effectiveness of the present IR method.

Bridge inspection is a pathway to bridge condition rating assessment, and is an essential element of any bridge management system (BMS). The success of a BMS is highly dependent on the quality of bridge inspection outcomes and accurate estimation of future bridge condition ratings. However, existing visual bridge inspection methods suffer several limitations due to human subjective judgment. In order to minimise such limitations, a feasibility study has been performed to enhance the current visual inspection method using optical image processing techniques. However, the accuracy of the inspection outcomes still requires further improvement. This paper proposes an automatic bridge inspection approach employing wavelet-based image features along with support vector machines (SVM) for automatic detection of cracks in bridge images. A two-stage approach is followed, in the first stage, a decision is made as whether an image should undergo a pre-processing step (depending on image characteristics); in the second stage, wavelet features are extracted from the image using a sliding window texture analysis-based technique. Consequently, an average accuracy of 92% (effect of training image types on accuracy) is obtained even when undertaking experiments with noisy and complex bridge images.

Traditional bridge inspection methods, such as visual inspections and basic non-destructive tests, remain foundational but face limitations due to labor intensity and dependency on inspector expertise, often resulting in inconsistencies. Machine learning (ML) tools offer transformative solutions by enabling the analysis of large datasets to enhance damage detection accuracy, optimize maintenance schedules, and improve resource allocation. The objective of this study was to explore the integration of advanced inspection tools, such as Unmanned Aerial Vehicles (UAVs), digital imaging, and fiber optic sensors, with ML models to enhance data accuracy, decision-making, and inspection rating efficiency. A state-of-the-art review is conducted in this study on the application of ML techniques in bridge inspection and maintenance, encompassing 60 articles and reports published in the last decade. The results of this study show that ML’s ability to integrate advanced inspection tools to improve data accuracy, decision-making, and inspection efficiency. However, the existing challenges persist in data quality, model generalization, and the need for standardization approaches across diverse bridge conditions. The review provides insights into current methodologies, benefits, limitations, and future directions, emphasizing ML’s pivotal role in modernizing bridge health monitoring for a safer and more sustainable infrastructure network. The findings of this study are expected to assist transportation agencies in planning and operating predictive maintenance strategies, focusing on damage detection, maintenance optimization, and resource allocation.

Many bridges and other structures worldwide present a lack of maintenance or a need for rehabilitation. The first step in the rehabilitation process is to perform a bridge inspection to know the bridge′s current state. Routine bridge inspections are usually based only on visual recognition. In this paper, a methodology for bridge inspections in communication routes using images acquired by unmanned aerial vehicle (UAV) flights is proposed. This provides access to the upper parts of the structure safely and without traffic disruptions. Then, a standardized and systematized novel image acquisition protocol is applied for data acquisition. Afterwards, the images are studied by civil engineers for damage identification and description. Then, specific structural inspection forms are completed using the acquired information. Recommendations about the need of new and more detailed inspections should be included at this stage when needed. The suggested methodology was tested on two railway bridges in France. Image acquisition of these structures was performed using an UAV for its ability to provide an expert assessment of the damage level. The main advantage of this method is that it makes it possible to safely accurately identify diverse damages in structures without the need for a specialised engineer to go to the site. Moreover, the videos can be watched by as many engineers as needed with no personal movement. The main objective of this work is to describe the systematized methodology for the development of bridge inspection tasks using a UAV system. According to this proposal, the in situ inspection by a specialised engineer is replaced by images and videos obtained from an UAV flight by a trained flight operator. To this aim, a systematized image/videos acquisition method is defined for the study of the morphology and typology of the structural elements of the inspected bridges. Additionally, specific inspection forms are proposed for every type of structural element. The recorded information will allow structural engineers to perform a postanalysis of the damage affecting the bridges and to evaluate the subsequent recommendations.

The integration of artificial intelligence (AI) in the form of large language models (LLMs) and generative models into clinical practice has progressed ahead of metrics available to measure their performance in real-world settings. Traditional benchmarks such as area under the receiver operating characteristic curve or bilingual evaluation understudy (BLEU) scores are inadequate to meet clinical nuance, patient safety, explainability, and workflow integration. This scoping review maps the evolving landscape of clinical AI evaluation, combining academic and industry architectures, including clinical risk evaluation of LLMs for hallucination and omission (CREOLA), hospital deployments, and radiological tool reviews. We explore stakeholder tensions between academia, business viability, regulation, and frontline usability, and reveal how these perceptions build competing evaluation imperatives. In particular, we highlight the novel challenges created by generative models: hallucination, omission, narrative incoherence, and epistemic misalignment. The current paper elucidates that a strategy of layered, stakeholder-engaged design needs to integrate risk stratification, contextual awareness, and continuous postdeployment surveillance. Equity, interpretability, and clinician trust are not thought of as footnotes, but as central columns upon which evaluation is built. This review offers a synthesizing overview of how health systems, developers, and regulators can coconstruct adaptive and ethically grounded evaluation frameworks, ensuring that AI tools enhance, rather than erode, clinical judgment, patient safety, and health equity in real-world care.

I served as the guest editor of the July/August 2010 issue of the Journal of Bridge Engineering. During the last 2 years, partly because of the failure of the I-35 Bridge inMinnesota in 2007 and other incidents, there has been significant focus and healthy discussion among bridge infrastructure stakeholders, including several elected leaders, in using nondestructive methodologies in assisting bridge inspection, evaluation, and maintenance. At present, nondestructive evaluation and testing (NDE/NDT) methods, such as ground penetrating radar and infrared thermography, are being commonly used by most transportation agencies. Several federal and state agencies are funding significant resources in developing and researching nondestructive test methods for quantitative evaluation of bridge infrastructure to augment visual inspection data. This special issue presents some of the recent advances and applications of nondestructive test methods and structural health monitoring for bridge evaluation and management. A majority of bridge failures are attributed to scour damage that is hard to detect in real time and appropriate preventivemeasures that are hard to apply when scour damage is detected. Hence, scour monitoring is an important topic for transportation owners, especially during high-flood events and in coastal areas. The technical paper “VibrationBased Method and Sensor for Monitoring of Bridge Scour” by Zarafshan et al. introduces a new fiber-optic Bragg grating (FBG) scour sensor. The scour-depth detection is based on the inverse relationship between the fundamental frequency and the length of the sensor rod embedded in the riverbed. This paper describes development of the theoretical basis for the sensor, computational methodology for detection of the riverbed foundation properties, laboratory and small-scale field verification tests, and installation and remote monitoring of scour in a multispan scour critical bridge in Illinois. Cable-stayed bridges are increasingly popular in the United States because of their elegance and the advantages they offer for relatively long spans. When these unique bridges are built, several design assumptions are made and may have to be verified after their construction. A cable-stayed bridge was recently constructed by the Ohio Department of Transportation. Following recent trends, the stays at this bridge were built without the use of grout for the purposes of inspection and possible replacement in the future. The bridge incorporated measures put forth to mitigate stay motion. The technical paper “Cable-Stayed Bridges: Case Study for Ambient Vibration-Based Cable Tension Estimation” by Kangas et al. presents experiments that were performed to determine the viability of using traditional vibration techniques that assume an integral sheath to estimate cable tension with this new configuration. Acoustic emission (AE)-based techniques have been widely used for detecting breaks in suspension and cable-stayed bridges. The technical paper “Detection of the Presence of BrokenWires in Cables by Acoustic Emission Inspection” by Zejli et al. studies AE techniques to detect the presence and location of broken wires in anchorages. Structural health monitoring (SHM) is becoming popular in bridge monitoring. System configuration is very important to effectively monitor performance parameters of interest. The technical paper “Measurement System Configuration for Damage Identification of Continuously Monitored Structures” by Laory et al. discusses a systematic approach to determine the appropriate number and location of sensors to configure measurement systems in which static measurement data are interpreted for damage detection using model-free (nonphysics-based) methods. A railway truss bridge in Zangenberg, Germany, is used as a case study to illustrate the applicability of this proposed approach. The technical paper “Approach to Reduce the Limitations of Modal Identification in Damage Detection Using Limited Field Data for Nondestructive Structural Health Monitoring of a Cable-Stayed Concrete Bridge” by Ismail et al. proposes a technique to reduce the limitations of modal identification in damage detection using reduced field data for nondestructive SHM of a cable-stayed concrete bridge. Use ofmonitoring systems for bridgemaintenance is also gaining popularity. The technical note “Automated Ice Inference and Monitoring on the Veterans’Glass City Skyway Bridge” by Kumpf et al. presents an ice inference system installed on Veterans’ Glass City Skyway (VGCS) Bridge to assist the Ohio Department of Transportation in managing the response to icing events. Variation in modal properties because of the environment has to be studied before they can be correlated to possible structural damage. If variation in modal properties, caused by temperature variations to which the structure is subjected, is less than the variation caused by the damage of interest, thismethod cannot be reliably used for damage detection. Hence, the technical note “Effect of Temperature on Daily Modal Variability of a Steel-Concrete Composite Bridge” by Mosavi et al. investigated the effect of temperature variations onmodal characteristics of a two-span, steel-concrete composite bridge in North Carolina, and addresses the extent and reason of the daily changes observed in its dynamic properties. The deck is a major component of a bridge structure and is in direct contact with live traffic, subjected to inclement weather, and deicing salts are directly applied over it. Deck performance directly affects the durability and life-cycle costs of a bridge. The technical paper “In-Service Condition Assessment of Bridge Deck Using Long-Term Monitoring Data of Strain Response” by Ni et al. focuses on obtaining information of interest (such as peak stress distribution and dynamic internal forces) to structural engineers using data from the instrumented Tsing Ma Bridge. Inspection and evaluation considerations are very important from planning and design stages to preserve the structural health and durability while minimizing life-cycle costs. The technical paper “Damage Evaluation for Concrete Bridge Deck byMeans of StressWave Techniques” by Shiotani et al. focuses on detection of the fatigue damage of concrete bridge decks utilizing propagation of stress waves. This experimental study concludes that by using sparsely arrayed AE technology, global integrity of bridge decks could be carried out so that local NDE methods can be employed for further investigation of areas of interest. Use of ground penetrating radar, impact echo, and infrared thermography based techniques are widely used by bridge owners to augment visual inspection data from bridge decks for informed

Abstract. Bridge inspection is a critical task in infrastructure management and is facing unprecedented challenges after a series of bridge failures. The prevailing visual inspection was insufficient in providing reliable and quantitative bridge information although a systematic quality management framework was built to ensure visual bridge inspection data quality to minimize errors during the inspection process. The LiDAR based remote sensing is recommended as an effective tool in overcoming some of the disadvantages of visual inspection. In order to evaluate the potential of applying this technology in bridge inspection, some of the error sources in LiDAR based bridge inspection are analysed. The scanning angle variance in field data collection and the different algorithm design in scanning data processing are the found factors that will introduce errors into inspection results. Besides studying the errors sources, advanced considerations should be placed on improving the inspection data quality, and statistical analysis might be employed to evaluate inspection operation process that contains a series of uncertain factors in the future. Overall, the development of a reliable bridge inspection system requires not only the improvement of data processing algorithms, but also systematic considerations to mitigate possible errors in the entire inspection workflow. If LiDAR or some other technology can be accepted as a supplement for visual inspection, the current quality management framework will be modified or redesigned, and this would be as urgent as the refine of inspection techniques.

Virtual reality visualisation of automatic crack detection for bridge inspection from 3D digital twin generated by UAV photogrammetry

Bridge inspectors make handwritten records during field inspection. The major drawback of handwriting is that it exposes the inspectors to danger when they are making inspections in high places. In cold, snowy regions, sometimes they also have to work under slippery, freezing conditions. The subsequent input of handwritten data into computers is complex and laborious, making it prone to human error. To address these issues, we developed the Field Inspection Recording System (FIRSt). The system compiles audio data and image data to produce bridge inspection reports, thereby eliminating the need for handwriting on paper and reducing the risk that inspectors will meet with accidents. In addition, converting the audio data directly into text data also avoids transcription errors and reduces the workload. The system affords improved safety, accuracy and efficiency of inspection. It has been put to use in bridge inspections.

Recently, with the deterioration of bridge facilities, demand has arisen for a method to inspect many bridges efficiently. One proposed bridge inspection method involves observation and inspection of cracks on undersides of bridges using a video camera mounted on an unmanned aerial vehicle (UAV) that flies under the bridges. There is an option to have a pilot operate the UAV, but it is desirable to have the UAV fly autonomously when efficiency of inspection is considered. Though there is a method using GPS for autonomous flight control of UAVs, there are many cases in which GPS cannot be utilized under bridges, and a new method is required for autonomous flight control in such places. The authors have already shown that autonomous flight control of UAVs can be achieved within the range of a monocular camera image by measuring the position of a UAV using camera images. However, since the flight range is bounded by the monocular camera image, it is necessary to move the camera position to fly the UAV autonomously in a wider space. In this paper, it is shown that a UAV can achieve autonomous flight control in wider spaces by constructing a single coordinate system for a combination of two camera images. In addition, considering that various measuring instruments might be mounted on a UAV, an adaptive control method capable of obtaining good control performance without changing the design parameters of the controllers should be applied. This method is useful for maintaining control performance when the total weight of the UAV changes. To show the effectiveness of our proposed method, we give an appropriate practical flight target orbit and present its experimental results.