In Japan, maintenance of existing bridges has become increasingly important due to the increase of aging bridges. Inspection technicians regularly visit bridge locations to conduct visual inspections of bridge damage to ensure that the damage is recognized and recorded. However, the results of visual inspections for the same location vary depending on the experience of the inspectors. In developing countries abroad, the maintenance and management sector has not been prioritized and there are deficiencies in the number, experience and capacity of bridge inspectors. With the spread of DX technology in the global infrastructure sector, there is a growing expectation for more advanced, efficient and cost-effective training of maintenance and management personnel using BIM/CIM. In this study, a three-dimensional model was created using the SfM methodology, and a number of Chinese university students were educated to conduct online inspection. Finally, questionnaire and discussion were conducted to examine the validity of maintenance and management talent training. The results showed that the method was effective.

BIM/CIM is being promoted in the construction field.While the establishment of effective utilization methods for 3D point cloud data contributes greatly to the promotion of BIM/CIM, handling large volumes of point cloud information presents many challenges, and efficient processing is required while controlling the computational load.In this study, we used point cloud segmentation, a technique for classifying point clouds using deep learning (Deep Learning), a type of AI (Artificial Intelligence) technology.Based on practical needs, modeling was conducted for use in the fields of river structure design and maintenance.The actual modeling flow was presented using real data, and it was shown that 3D modeling has the potential to contribute to the sophistication and labor saving by efficiently modeling without losing necessary information.

In this paper, we examine the presence of subsurface objects and the detection of four types of objects from ground penetrating radar (GPR) images obtained by a webcam using YOLOv7 in real-time. In this study, four types of objects (styrofoam, wood, concrete, and aluminum) were buried in a sandbox, and 64 radar images were generated using 2600 MHz radar for learning and detection by YOLOv7. The 64 images were augmented to 192 using the Cutout method, of which 168 were used for training and 24 for validation. In addition to the 192 images for training and verification, 64 test images were generated and cross-validated ten times to evaluate real-time detection by the webcam. The result is an accuracy of about 89% for the presence of objects and a detection rate of 39-83% for four types of objects.

In response to the spread of COVID-19 from 2020 to 2022, efforts have been made worldwide to prevent the spread of the disease. A state of emergency has been declared four times in Osaka City to prevent such an epidemic. In this study, a time-series analysis was conducted using data on the flow of people in Osaka City to analyze the effects of emergency declarations. The effectiveness of the emergency declaration was verified by comparing predicted and observed changes in outings before and after each declaration. Furthermore, by mapping the flow of people in each area, the spatial reality of the change in the flow of people from dense to dispersed in the central area of Osaka City could be observed. The results of these analyses quantitatively confirmed that people initially tended to refrain from moving in response to the declaration but that the restraining effect of the declaration decreased as the declaration was repeatedly issued.

In the construction industry, we are promoting DX (digital transformation) of site management to improve productivity and working style reform as a response to the shortage of workers and long working hours. DX not only digitizes on-site information, but also makes advanced use of that data, supports decision making, and simplifies workflow. The authors believe that the digital twin is one of the key solutions of realizing DX. This paper reports on the development details and the use cases.

In this study, we estimated the type, the number, the areas, the weights of floating plastic litters in the images by using YOLO and DeepSort. The images of floating plastic litters were taken in an open channel by using visible and infrared cameras. The areas of them are estimated by using the Bounding Boxes by YOLO considering the litter’s rotation. The weights of them are estimated by the relationship between the areas of the plastic products and the weights of that. As a result, the accuracy of type detection was 0.83, F-measure was about 0.8 for number counting by DeepSort model and R² (coefficient of determination) was about 0.8 for the area estimation. For the weight estimation of each plastic types, the accuracy was lower compared to the estimation of the areas. However, the MAPE_m errors in the total weight of bottles and the packages were small (10 ~ 30 %). Therefore, we conclude that this proposed method is applicable for estimating the weights of floating bottles and packages, in the conditions of the experiment.

In Japan, the declining population, falling birthrate, and aging population in rural areas are posing the challenge of how to maintain and revitalize the functions of local communities. In order to develop the tourism industry, it is necessary to understand the points to be improved in tourist attractions as perceived by tourists, and to plan and implement new measures accordingly. This study collects word-of-mouth data on tourist spots from travel information websites, which are among the big tourist data accumulated by the spread of the Internet in recent years, and extracts sentiments that suggest negative feelings of touristsfrom the word-of-mouth data by performing sentiment analysis using deep learning. Based on the evaluations obtained through interviews with tourism researchers, we discussed the significance of the system and points to be improved.

Fukui Prefecture will become a region along the Hokuriku Shinkansen line when the line between Kanazawa and Tsuruga begins service in the spring of 2024. Fukui Prefecture is expected to see a significant increase in the number of people who travel not only to the Tokyo metropolitan area but also to the Kansai region and other regions due to the time-saving effect of the Tsuruga extension. Therefore, it is anticipated that the human flow trends in the Hokuriku region will change significantly after the spring of 2024. In this study, we examine the competition between railroads and airlines that is expected to arise as a result of changes in human flow trends and clarify the human flow trends at Fukui Station and Kanazawa Station, which are Hokuriku Shinkansen stations, and at Komatsu Airport, which is the "air gateway" to the Hokuriku region, before the opening of the Hokuriku Shinkansen Tsuruga Station. This will contribute to the further development of the Hokuriku region after the Tsuruga extension of the Hokuriku Shinkansen begins operations. This study clarified the relationship between the means of transportation of visitors to the Hokuriku region and the attributes of the users and their places of residence.

In non-targeted analysis, which has become mainstream in recent years in screening of chemical substances contamination in environmental samples, interpretation of measurement data is important. We have proposed a method to discriminate excitation-emission matrix (EEM) image by machine learning as a preliminary screening of pesticides contamination in environmental samples. In this work, we tried the feasibility of discrimination by AI using EEM image data from river water samples that were spiking pesticides and non-spiking. In addition, we examined the optimality of image data as training data. As a result, it was shown that this method can be used for preliminary screening of contaminated pesticides detection, and even if the amount of information is increased by precisely measuring the EEM image, it is not necessarily possible to improve the accuracy of judgment by AI using AlexNet.

Due to the current problems of deterioration of reinforced concrete structures and the shortage and population aging of engineers, there is a demand for methods that allow non-skilled engineers to quantitatively inspect reinforced concrete structures. In this study, by using electromagnetic wave radar, which is relatively easy to detect reinforcing bars and internal cracks, the estimation of microcracking in concrete at the early stage of deterioration was investigated using machine learning. As a result of the study, using the image extracted only the part of reinforcement from data measured by the electromagnetic radar, it was shown that it is possible to estimate cracks with about 0.04nnn width by learning and judging those images with the convolutional neural network.

This study proposes a dam inflow forecasting method that uses ensemble prediction for precipitation to reflect prediction uncertainty. The need to improve the accuracy of dam inflow prediction for effective dam operations has increased owing to large floods that have been occurring frequently across Japan in recent years. This study focuses on a dam in Hokkaido, Japan, which has been prone to floods in recent years. Elastic Net, which is a sparse modeling method, was used to predict inflows. Meso-scale Ensemble Prediction System (MEPS), the ensemble prediction for precipitation, was introduced as an input to the model. This was compared against predictions using the storage prediction method to evaluate accuracy. The results suggest that introducing MEPS into Elastic Net can provide accurate forecasts with safe results from a flood control perspective.

With the development and advancement of AI technology, research on the application to the field of infrastructure maintenance is actively progressing. Many of these studies focus on the development of learning theories that consider the characteristics of images obtained in infrastructure maintenance. The effectiveness of AI has been demonstrated in various tasks such as crack detection, classification of defect types, and estimation of degradation levels. On the other hand, to truly enhance the operational efficiency through AI, it is necessary to construct AI systems considering the practical business. Furthermore, to improve AI and continuously utilize AI, it is necessary to acquire images suitable for AI development in operations. Therefore, this paper introduces the learning theories that have been developed for images obtained in infrastructure maintenance, previous research on AI with essential function for the practical business and the authors’ idea on efficient image acquisition.

The spread of novel coronavirus infection is now coming to an end in Japan, and the number of foreign tourists, which had declined significantly since 2020, is on a recovery trend. Ishikawa Prefecture is no exception, and the number of foreign tourists is increasing again. Furthermore, it has been decided that large-scale tourism measures will be implemented in Hokuriku when the Hokuriku Shinkansen bullet train opens in Tsuruga in the fall of 2024. Against this background, the purpose of this study is to clarify the spatial relationship between the distribution of stays of foreign visitors and tourist spots in Kanazawa City. In this analysis, land in the central city area of Kanazawa was clustered in each 500-meter square mesh according to the number of tourist spots, lodging facilities, and restaurants, and compared with the distribution of foreign visitors to Japan. As a result, we were able to identify popular areas for lodging, popular areas for tourist spots, and popular areas for food in Kanazawa City.

In the developing country of the Democratic Republic of Timor-Leste, numerous slope failures have occurred, reducing road capacity and severely reducing accessibility between urban and rural areas in a very large number of cases. In addition, the restoration process takes an enormous amount of time, and road capacity has not been improved. In fact, a survey conducted by the authors revealed that many citizens are dissatisfied with the decline in road capacity and would like to see it restored as soon as possible. In this study, we conducted a survival analysis of the willingness to pay for slope failure restoration among the citizens of the Democratic Republic of Timor-Leste. Through the analysis in this study, differences in willingness to pay by recovery period were clarified, and the willingness-to-pay model revealed that the level of dissatisfaction affects the amount of willingness to pay.

This study proposes a simple and useful system for measuring and evaluating cracks on road pavements, by updating the same type of system developed in our previous research. In the present system, an action camera (Hero10) is employed, since it can work for more than one hour with a battery and save images and GSP data simultaneously in the memory inside. With some auxiliary devices, it can easily be attached on the roof of a car and capture the same area of road pavements by adjusting those devices. The recorded images are segmentated by an advanced neural network Deeplab v3+, in order to distinguish asphalt pavements from the other objects. Then, each segmentated image is divided into small blocks and they are classified into eight categories according to the state of cracks by a well-known neural network ResNet-18. Finally, crack rates of the images are computed from the results of the classification with new step-wise weights and they are visualized as color contour on a map.