The use of ICT in snow plowing and removal operations is being promoted in snowy and cold regions due to an aging workforce and labor shortages. However, the digitalization of the snow removal patrols, which are used to determine the timing of snow removal, is still in preliminary stages. The vehicle-mounted mobile mapping system (MMS), which transforms the snow conditions of roads and shoulders into 3D point clouds, is useful in advancing the digitalization of snow removal patrols. This study develops an MMS integrated with a low-cost GNSS/INS to investigate the possibility of adopting an affordable MMS for snow removal patrols. The accuracy of map coordinates in MMS-generated point clouds in the post-boresight calibration has beenwas evaluated using control points in urban blocks. The map coordinate accuracy has been evaluated to be 0.085 m (3D) and 0.032 m (2D). Based on this result and the availability of inexpensive commercial LiDARs, an inexpensive MMS can be developed for measuring snow cover on roads and shoulders.

 Since social infrastructure, which was intensively developed during the high economic growth period, will deteriorate all at once in the future, maintenance and management of facilities will be an issue in the future, and it is important to understand the local situation through appropriate visualization of inspection results. The authors developed a method to grasp the position, direction, and size of deformation in relation to the entire facility by pasting inspection photographs taken close to the deformed part of the sabo dam onto a 3D model of the sabo dam. There was a problem in that inspection photographs needed to be post-processed using SfM/MVS processing (hereinafter referred to as SfM processing). Therefore, in this paper, we compared a 3D model generated by a 3D reconstruction application using a mobile terminal with a 3D model generated by SfM processing, and investigated a method for importing the 3D model into GIS.

 Radical measures are expected to reform the way of working at construction sites. One effective approach is to quantify the movements within the site and evaluate labor productivity. In this study, we first conducted a hearing survey with on-site engineers and systematically organized the focus points (where engineers are paying attention) in the five elements of construction management, QCDSE, based on the survey. We demonstrated that these focus points can be quantified as Key Performance Indicators (KPIs). Next, we focused on D (Delivery) and S (safety) in QCDSE and developed a system to visualize the operation rate of construction machinery and potential risks, applying it to a dam construction site. By evaluating the operational effectiveness of the system from the perspective of on-site engineers, we confirmed the usefulness of the developed construction management system, as it can streamline management tasks for engineers and extract previously overlooked events.

 In this study, we proposed a system that can instantly estimate the quantity, composition and spatial distribution of disaster waste that was generated by inundation of river water flooding. Because all developed functions and mandatory databases were implemented on the online Google Colab, system users could estimate disaster waste generation status by only uploading an inundation map. There were no troublesome environment setting of each local computer, users didn't have to spend time configuring their computers during the chaos of a disaster. All data in database was copyright-free, so there were no restrictions on its use. We applied the proposed system for 2 inundation events (Nagano city in 2019 and Hitoyoshi city in 2020) aiming at evaluation, it could be demonstrated the system could output reasonable disaster waste quantity, composition, and spatial distribution of disaster waste immediately.

 In road management during the snowy season, it is important to properly determine changes in snow depth on slopes next to roads. UAV-SfM surveying, which has been increasingly used in recent years, is an efficient method for measuring snow depth, but it has been pointed out that it is difficult to stably and permanently establish multiple ground reference points on snow-covered slopes, which is necessary to accurately superimpose the positions of point clouds during the snow-free and snow-covered seasons. In this study, we focused on the differences in positioning methods of UAVs and the number of ground reference points to be set up on slopes, and verified their measurement accuracy. As a result, it was confirmed that the measurement error of snow depth could be suppressed to +6.4 cm (8%), which is practicable for snow depth measurement, under the following conditions: (1) a UAV equipped with RTK is used, (2) set the ground pixel size to the equivalent of 1 cm when shooting, and (3) at least two ground reference points are installed for SfM analysis. This result indicates that the UAV-SfM surveying system can improve the efficiency of ground reference point placement during the snowy season.

 In recent years, the construction of new urban expressway routes required the application of unprecedented complex structures, such as significantly eccentric bridge piers, due to various constraints. When conducting load tests on such structures, it is anticipated that complex local buckling with twisting deformations, unlike anything seen before, may occur due to out-of-plane (bridge axis) loading. Therefore, traditional deformation measurement methods may not facilitate the prediction of deformation locations or the acquisition of diagonal deformations. In this paper, we report on the potential utility of measurement using a high-precision point cloud data acquisition scanner for out-of-plane loading experiments on highly eccentric concrete-filled steel bridge piers, with the aim of accurately capturing the complex local buckling shapes involving twisting deformations. This research investigation the possibilities for future applications.

 In current pedestrian traffic volume surveys, pedestrians passing through the surveyed cross section are generally counted manually, which limits the survey days and times. In recent years, there has been an increase in the number of surveys using video images taken by video cameras, but personal information and privacy must be taken into consideration. Therefore, LiDAR, which can measure the target pedestrian as a set of three-dimensional coordinate points, has been attracting attention. However, repetitive LiDAR cannot measure the measurement range exhaustively and is difficult to be applied to the survey. In this study, we conducted a pedestrian traffic survey using deep learning with point cloud data measured by non-iterative LiDAR, which can measure the measurement range exhaustively. The results of pedestrian counting showed that the correct response rate was 67.2% in the low case and 84.7% in the high case, indicating that the point cloud data measured by non-repeatable LiDAR has potential to be applied to pedestrian traffic volume surveys.

 With the rise of i-Construction in Japan, there has been an increase in opportunities to acquire point cloud data that captures the surface layers of the earth in three dimensions are increasing. Structure from Motion (SfM), which can produce point cloud data from multi-view images, is particularly noteworthy. However, because SfM determines corresponding points and self-positions based on the similarity of feature values derived from the RGB values of each pixel, it struggles to generate dense point cloud data for objects that have minimal color variation, such as a white wall. In previous research, we proposed a technique to enhance the density and accuracy of SfM point cloud data by projecting a pattern onto the object with a projector. However, this method introduced the issue of creating artificial unevenness that doesn't actually exist on the object. In this study, we devised an optimized pattern to generate accurate point cloud data and validated its efficacy with demonstration experiments on a bridge pier model. The results showed that, using our method, SfM can produce point cloud data with an error margin of less than 10 mm when compared to laser-based equipment.

 In Japan, AI is being incorporated into automobile traffic volume surveys to enhance efficiency. By recognizing license plates in these surveys, it's possible to measure vehicle flow, refining the accuracy of the survey. However, when the video footage is taken from a high vantage point on the road's shoulder, the license plates appear tilted. To correct this, the image needs to be adjusted to a frontal orientation. If the license plate and the vehicle have the same color, determining a reference point for the projective transformation becomes challenging. To address this issue, we developed a method that corrects the license plate to a front-facing orientation using a projective transformation, based on the position of the fourth digit of the serial number—a consistent feature on all license plates. Our experimental results indicated successful character extraction from 479 out of 500 images. We aim to apply this technique in future traffic flow studies to identify the same vehicle at multiple locations.

 Counting the number of people in a crowd at an event or during a disaster is important to prevent accidents from occurring. In recent years, various methods have been developed to count the number of people using deep learning. However, it is difficult to automatically select and apply the most appropriate counting method, because video images of crowds are taken under various conditions. In this study, we focused on the differences in the trends of the detection results of head counting methods and devised a method to estimate the number of people using regression analysis. As a result of the verification, it was found that the regression analysis can reduce extreme false positives and omissions and estimate the number of people more accurately by complementing the weaknesses of each head counting method.

 This study shows that we implemented two alternative methods, instead of a normal Bayesian inference method, into Long, Short-Term Memory (LSTM) to visualize model-related uncertainties. This model is called Bayesian neural networks (BNN), which was applied to predict water levels in a reservoir used for drainage management in a lowland area. In BNN, we used two alternative methods: Monte Carlo dropout (MC Dropout), which is a method of randomly switching nodes in the network, and Stochastic Gradient Langevin Dynamics (SGLD), which is a sampling method. The BNN predictions of MC Dropout with the best dropout rate (=0.3) were performed by 10% improvement of accuracy when compared to the conventional model in short lead time for the drainage period during the largest flood. SGLD-based BNN had equivalent results to the BNN-MC Dropout. Comparing among the predictions of both methods, including credible intervals, MC Dropout-based BNN showed wider and smoother temporal distributions, especially near the peak water level.

 When subject to excessive loads in the context of excavation construction, earth retaining walls can undergo deformation that is challenging to assess visually, potentially resulting in accidents involving wall collapses. Hence, previous research has explored installing sensors within the retaining walls and using surface targets in conjunction with TS to measure deformation quantitatively. Nevertheless, these methods have presented significant challenges, including high costs, extensive labor requirements, and limitations in comprehensively assessing deformation. Therefore, we focused on cost-effective LiDAR technology, which offers the capability to provide real-time, comprehensive 3D data for measurement targets. We proposed a method for detecting deformations in earth retaining walls using LiDAR, representing a non-repetitive measurement approach. The experimental results demonstrated the effectiveness of the proposed method in detecting surface deformations in earth retaining walls. In the future, our objectives encompass applying this method to practice construction sites, identifying and resolving implementation challenges, and refining the proposed technique.

 Due to climate change, the frequency of intense rainfall will increase, and there are concerns that more rainwater will flow into roads in a shorter amount of time than before. Prediction of road flooding can be useful for determining safe routes for road users, and also for road managers to make traffic regulations more efficient. In this study, we used flooding sensors, which are battery operated radio device and lower cost and space-saving than continuous water level gauges, to predict road flooding. We aimed for practical road flooding prediction by focusing on flooding sensor and rainfall with a simple method. Since the water level information that can be obtained from water detectors is limited, this paper proposed a simplified model for a water level prediction. In a experiment, we applied this method to the rainfall from Typhoon No. 2 in June 2023 and information from flooding sensors at two locations in Saitama City. The predictions captured the general shape of water level changes. In order to understand the issues involved in introducing road flooding prediction into society, we developed an information system that visualizes road flooding prediction, and used this to obtain practical opinions from road administrators.

 Road administrators conduct road surface condition surveys and visual inspections to monitor the condition of road pavements. However, they face challenges in terms of labor and cost. Therefore, technology that uses deep learning to analyze video images has been attracting attention. These images of road pavements are captured by drive recorders and smartphones. In existing research, the accuracy of the training data is low, and shadows on the road surface reduce the detection accuracy of cracks.

 In this study, we developed a deep learning model designed based on specific characteristics and methodologies derived from the results of a road surface condition survey. Additionally, we devised a simple method to calculate crack rates by removing shadows from 4K-resolution video images using machine learning. Through demonstration experiments, we found that the crack ratio from the road surface condition survey results had an accuracy of approximately 60%. This is equivalent to the accuracy of visual inspections, suggesting that cracks in road pavements can be easily inspected.

 Recently, the Ministry of Land, Infrastructure, Transport and Tourism has advocated for the application of BIM/CIM principles, to promote design project on digitalization and utilization of 3D models. Such digital transformation Such digital transformation allows stakeholders to be more able to manage data sharing and utilization, which aims to improve productivity in design, construction, maintenance, and management. However, creating 3D models for existing structures on bridges requires measurement data, specialized knowledge, and substantial effort. Previous researches have proposed methods to automatically generate parametric models based on point cloud data and bridge component template models using genetic algorithms. However, these methods need significant processing time. Therefore, in this research, a more efficient method is proposed to generate parametric models by employing deep learning to estimate edge data on bridge component point cloud data, after which wireframe models are subsequently created. The proposed method is possible to generate parametric model with higher accuracy and speed compared to existing methods.

 Governments use the green cover ratio and the greening ratio of roadside areas as indicators for evaluating the green cover of residential areas. However, these indexes are based only on the existence of vegetation and do not fully reflect the uneven distribution of vegetation or the diversity of its growth status. In this study, effects are examined using high-resolution satellite images, which are expected to be used for maintenance and management, including the continuity of vegetation as an indicator. As a result, an index value reflecting the vegetation condition was proposed by relating the gravity model for relating vegetation cells on the analyzed images to the NDVI value indicating the vegetation activity of the cell itself, and its effectiveness was confirmed in comparison with previous index values. The proposed index value was used to evaluate the vegetation environment in Senri New Town in Osaka Prefecture by dividing the streets into those including roadside trees and those not including roadside trees in each district, and its effectiveness was discussed.

 Although large-scale reconstruction projects were carried out over a wide area in the areas affected by the Great East Japan Earthquake, the amount and spatial distribution of anthropogenic disturbance caused by these reconstruction projects are not clear. In this study, we estimated the spatial distribution of topographic changes caused by reconstruction projects based on the difference in elevation between the DEM immediately after the earthquake and the current DEM. As a result, it was visualized that human-induced landform changes such as elevation changes due to residential land development and raising, and cut and fill due to road construction.The amount of landforms were estimated to be 629 million m³ in Iwate Prefecture and 596 million m³ in Miyagi Prefecture, indicating that the amount of alteration was large in the area from Miyako City to Higashi-matsushima City. The amount of landforms alteration per area using the zoning before and after the earthquake was calculated, and it was found that residential land development was responsible for the large amount of alteration intensity. Although the amount of landform alteration is small compared to other human-caused landform alterations such as open-pit mining and sediment extraction, it is considered to be a large-scale landform alteration because the reconstruction project was implemented over a wide area.

 This paper presents a digitalization of hammer testing in tunnel inspection from movies in which the activity is recorded. The proposed method first extracts keyframe images at the moment of hammer-hit on surface through the acoustic analysis of hammering sound in the input movies. For each keyframe, our method computes the hammered position, and converts it to the position in the coordinate system of a base image known as labframe image using homography transformation of the two images. The result is obtained by the integration of hammered positions, associated times and sounds, and it makes possible additional analyses including distribution analysis and evidence of judgement. This paper also demonstrates applications to a skill evaluation of tunnel inspectors.

 In this study, we develop a system that uses Mixed Reality (MR) to support residents in making disasters personal by facilitating the construction of disaster images of their own town. Using this system, we will reproduce disaster situations based on the assumptions of hazard maps and the memories of disaster victims. The system also allows multiple residents to share the recreated images of the disaster so that they can communicate with each other while recreating the disaster. In this study, we conducted an evaluation experiment with residents and verified the effectiveness of the system through questionnaires and opinion exchange meetings. As a result, we confirmed the effectiveness of the MR-based system in terms of imagining disasters and making disasters personal.

 In recent years, with the accelerated global uptake of renewable energy, championing wind power generation projects in our country has become crucial. Achieving a consensus among stakeholders and local residents is essential. A primary challenge lies in forecasting and visualizing the environ-mental changes instigated by these projects. Concurrently, advancements in digital technologies, exemplified by PLATEAU buildings, have heightened expectations for Digital Transformation (DX). This paper introduces a method of project presentation harnessing Metaverse visualization techniques. To streamline consensus-building, we suggest a system adept at instantaneously updating vast quantities of virtual space information within the Metaverse. We particularly advocate for the use of these techniques in predicting facility-induced shadows and elucidating construction methodologies. Additionally, through questionnaire-based consultations with project operators, we underscore the efficacy of these visualization methods.

 In this study, in order to examine the use of centimeter-level positioning augmentation service (CLAS) by QZSS. in surveying, we verified positioning solutions after reinitialization when positioning at known points. We averaged the data from the CLAS fix solution acquisition start time to 180 seconds (epoch) over several time periods, and evaluated the difference from the known point coordinates of plane rectangular coordinates X, Y, and altitude H, and the RMS error. As a result of the experiment, the maximum difference in the X and Y coordinates of CLAS is approximately 75% and 67% within the permissible value of ±2 cm, and the maximum percentage of the difference in elevation H is within ±3 cm, which is the permissible value. It became 30%. For the RMS error, the 10 epoch average value is lower than the 1 epoch average value in the X and Y coordinates, but it increases at the 20 epoch average value and 30 epoch average value, and gradually decreases after the 60 epoch average value, and the 180 epoch average value became the minimum.

 Road managers conduct visual inspections and road condition surveys to identify road pavement damage, but there is a shortage of manpower and cost issues. To solve this problem, a method for detecting rut excavation by applying deep learning image analysis to video images captured by an in-vehicle camera has attracted attention. Existing studies have demonstrated methods using single-lens reflex cameras and video cameras, etc. However, detection using easily available video images from drive recorders and smartphones is expected to accelerate the development of infrastructure DX.

 In this study, we developed a method for detecting rut excavation using image segmentation based on deep learning by correcting the gamma value representing the luminance of 4K video images captured with a drive recorder and a smartphone. As a result, Mask R-CNN and YOLOv8 were able to construct models with fewer false positives and omissions, respectively, suggesting useful implications for practical applications.

 In recent years, measurement of 3D point clouds using Mobile Mapping Systems (MMS) has become popular. There are two types of laser scanner measurement methods: Time-of-Flight (ToF) and Phase Shift (PS). Existing studies have shown that the PS laser scanner is more accurate than the ToF. On the other hand, the PS method sometimes generates peculiar noise when the reflection intensity is high, such as in the case of a sign, and the shape of the sign may not be accurately determined. Therefore, we used an MMS equipped with a PS laser scanner to clarify the noise generated by the reflection and material conditions at each angle of incidence of the laser on target plates of different colors and surface materials, and investigated how to deal with this noise. As a result, it was shown that the measurement at an angle of incidence 60° was suit for scanning for both paint-coated and reflective materials.

 The development of heat indices based on quantitative assessments of heat stroke risk has become an important source of information for addressing the contemporary issues of increased risk of heat stroke due to rising global temperatures and the heat island effect. Heat indices for urban areas provided by the Ministry of the Environment are generally available on the Web, but this information is limited for individual citizens to understand the risk of heat stroke in their immediate living and working environments. If the difference and distribution of the risk between sunny and shaded areas, as well as changes with time in the same hot environment, it would be easier to plan individual activities and tasks. In this study, the heat index is estimated pixel by pixel based on computer graphics rendered with realistic shadows by global illumination, using 3D data and weather data, which are public information infrastructure, to calculate sunlight conditions for specific sites and local environments. We developed a system that visualizes the heat index distribution in real time for sequential changes in the heat environment and distributes it on the Web by implementing the system with a game engine.

 In the after-coronas, there is still a high demand for information on the degree of human crowding because of the importance of continuing infection control measures for the elderly, who are at high risk of serious illness. The technology to detect people using monitoring cameras is costly in terms of the number of cameras installed in wide areas, and has limitations in terms of coverage and installation locations. In this study, we propose a method for displaying and updating the distribution of people using images obtained by recording live streaming video from an omnidirectional camera while moving. The human distribution was qualitatively visualized by obtaining and mapping the positions of people on a floor map of a facility using alignment and machine-learning-based human detection and tracking processes. As a result, the system was able to display and update the distribution of people online in parallel, even over a wide area, demonstrating the usefulness of the system in practical use.

 In the maintenance and management of highway structures, various platforms have been proposed to spatially integrate and utilize various types of information on a 3D model. By building a platform, information on different structures such as bridges and pavements can be managed centrally using location information. On the other hand, it is difficult to construct a platform for complex structures such as special bridges because it is difficult to create a detailed 3D model, and there are also distortions and lack of location information in the inspection records. In this study, we proposed a new 3D modeling method for special bridges that changes the level of detailing for each member, and a method of integrating inspection and repair information and models by regularizing location information, and constructed an information integration platform suitable for use in maintenance and management. The visualization and spatio-temporal analysis of the information obtained from the platform were conducted, and the usefulness of the platform for maintenance management was clarified.