The setting of tolerances for geometric characteristics of 3D models can be more accurately described by using geometric tolerance focusing on the geometric characteristics, rather than dimensional tolerance focusing on dimensions.
In this research, we consider to apply geometric tolerance annotations to 3D design models of the piers, we realized that we could apply geometric tolerance and meet the criteria of civil engineering construction management standard. The reviewed items are bridge framework, prefabricated pile work (steel pipe soil cement pile), reinforcement work (assembly).The new methods make possible to improve precision of civil engineering structures as accurate information is transmitted. It is eliminated ambiguity and misinterpretation compared with representation by conventional dimensions.
As the main task for concrete placing uses manual operation, such as the use of specialized equipment, it is very important to pass down the empirical knowledge of skilled technicians to their juniors. With the development of the basic technology of skill succession in concrete placing as the final goal, the authors have proved, through hearing surveys with technicians and so on, that the action of inserting a vibrator has a significant influence on empirical knowledge. A method of measuring the position and depth of vibrator insertion was also devised, and its usefulness verified. However, since it was sometimes impossible to measure the insertion depth at the actual site, improvement in the measuring method is mentioned as a future challenge.
In this study, a method of measuring the insertion depth of a vibrator using a gimbal mounted camera was devised. Measurement experiments were conducted based on the devised method, and useful findings were obtained.
In Japan, there are 111 active volcanoes that account for about 7% of the world. Once a volcano erupts, devastating damage occurs due to eruption events such as volcanic cinders, pyroclastic flows and debris flows. Therefore, it is important to promptly detect signs of eruption and take countermeasures through regular observation and monitoring of active volcanoes.
In this study, we considered the method using deep learning of AI technology to improve the efficiency of active volcano monitoring. Specifically, by using CNN(Convolutional Neural Network) of the deep learning model, a model that removes noise such as clouds and fog that hinders volcano monitoring and a model that detects eruption events such as smoke of volcano and debris flows were constructed. The target volcano was Yakedake, one of the 50 active volcanoes that the Japan Meteorological Agency is constantly monitoring. As a result, it was shown that deep learning could be an effective technique for improving the efficiency of active volcano monitoring.
Recently, an asset management system is an effective tool for maintaining to civil infrastructure. Asset management is established for a purpose of preventive maintenance composed of PDCA (Plan-Do-Check-Action) cycle. In Plan step, a plan of replace and repair is made from based on past inspection results. In Do step, the current state is evaluated from the inspection and repair performed to the infrastructure. In Check step, it analyzes the difference of state between initial predicted and current for investigating the cause. In Action step, re-plan is made from correcting to the cause. The preventive repair in the Do step has a role of significant for extending the life of the infrastructures.
In this study, a simple repair is introduced to the observation state corresponding to not need repair for preventive management in tunnel lighting facilities. Then, it estimates future prediction deterioration and repair costs to a ratio for repair amounts of observation state. It is decided the optimum repair ratio on indicator of balance for deterioration and repair costs from these results.
This paper is formulated the concrete methods. Firstly, health degree for deterioration indicator expressed as a numerical value from 0 to 100 is calculated by not micro model for each tunnel lighting facility but macro model for an aggregation of facilities in unit of a tunnel tube. Then, it is calculated the health degree and the repair costs corresponding to repair ratio by simplified dynamic macro model which proposed by authors in past. It is proposed the optimum repair ratio by the cost effectiveness of indicator for balance health degree and repair cost. Finally, it is explained validity and problem for performance having this model.
For the maintenance and management of roads in Japan, over 1.2 million km of road pavement is inspected for deterioration, and repairs are carried out according to the results.Generally, during road pavement inspections, visual inspections and road surface property surveys using high-precision measuring instrument laser technology are carried out. However, road surface surveys are often difficult to implement in a timely manner due to the high survey costs.
In this study, we devised an "effective road pavement inspection method" that narrows down pavement degradation candidates using commonly available car probe data, which was obtained in Fujisawa City, Kanagawa Prefecture. An inspection equivalent to a road surface property survey using a mobile measuring vehicle (MMS) was conducted, and the effectiveness of the proposed method was confirmed by comparison with the results of a conventional survey.
As a solution to congestion in an urban area caused by an excessive car society, the introduction of next-generation public transportation systems is being reviewed, and the introduction effect has been improved to facilitate consensus building with neighboring residents when introducing new transportation means. It is important to explain clearly. Therefore, we developed a 3DVR urban traffic flow modeling system using an integrated development environment of traffic flow micro simulator and game engine based on commercially available 3D city model data. And, using mobile VR that is cheap and easy to carry, we verified the applicability of VR content created by this system. As a result, we confirmed that VR content by this method can be experienced without lowering the display quality and processing load compared to the method using conventional CG modeling tools. In addition, it is expected that the man-hours and period for creating VR contents can be significantly reduced.
In recent years, the number of surveillance cameras installed in towns and commercial facilities has been increasing steadily. By identifying and tracking people captured by each camera, an improvement in the efficiency of criminal investigations or flow analysis within facilities can be expected. Moreover, applying cameras to construction sites enables operators working there to recognise their invasion into dangerous places with risks of falling or tumbling as well as recording near-misses with construction machinery, thus improving safety management. To realize these goals, it is necessary to implement the technology of automatic person identification.
Many existing studies have applied deep learning in recent years, which reports that face authentication, gait identification, and human identification lead to results with a higher precision than ever before. On construction sites, however, it may not be possible to apply the existing technology as operators’ clothes tend to be similar to each other and it is necessary to identify construction machines. In this study, we propose a method for human identification based on the convolutional neural network of deep learning with excellent image recognition, with special attention to drawing patterns on items that operators always wear on construction sites such as safety vests and helmets. Evaluation experiments were conducted to verify identification precision, and prove the applicability of the proposed method for the safety management of construction sites.
The surface appearance of weathering steel is significant factor for evaluation of soundness of weathering steel. The inspection of weathering steel is carried out by visual inspection method, however, it is difficult to judge the state of surface rust accurately. Therefore, in this study, the relationship between the surface appearance of weathering steel and its fractal dimension was investigated. The results showed that fractal dimension became higher when surface appearance of weathering steel is rougher, that is, fractal dimension related to the surface appearance. However, it was considered that rust evaluation by using fractal dimension is the part of the evaluation of soundness of weathering steel by inspection expert.
Initial Galileo service was launched in 2016, and the convenience of using Galileo is increasing in Japan, so it is expected that Galileo will be used in combination with GPS and QZSS in the future. This study was conducted to inspect the effects of kinematic positioning with combined GPS/QZSS and Galileo in term of Fix rate and RMS error, etc. in various surrounding conditions where the effects of multipaths, etc. are concerned. By comparing with the results of using GLONASS together, the significant effect of using Galileo in combination with GPS and QZSS under multipath environment was shown.
For efficient and effective maintenance of aging piers, the camera-mounted radio controlled boat was developed. This enabled the acquisition of 3D dense polygonal model with rich texture of the undersurface of the upper part of the pier by SfM/MVS process to mulpiple images. In this paper, for the effective use of 3D model in various stages of maintainanc process, we propose the method for automatically converting it to the textured simplified model. To create the simplified model, our method first extracts planar surfaces accurately each of which represents the surface portions of the constituting parts, such as slab, beam, and pile top. It also creates high resolution ortho images of each planar region along its boundaries, which can be mapped to their corresponding regions of the simplified model. Cracks and flakings can be detected on the ortho images, and the images can be stored with the model altogether. We demonstrate the effectiveness of the proposed method through various experiments.
At the construction site, as-built management is generally performed by taking pictures and comparing them with drawings or Building Information Modeling (BIM) models. Since this work is time-consuming and prone to human error, a more accurate and efficient method of capturing the progress is desired. The purpose of this research is to construct a system that can efficiently capture the progress of the construction by detecting each structural steel frame component such as a beam and a column under construction from images taken by a camera. First, we developed a Convolutional Neural Network (CNN) that could detect structural steel frame components under construction from images by fine-tuning the existing Object Detection and Segmentation CNNs. Next, we constructed a system that can capture each structural steel frame member from an image by integrating two constructed CNN models. Then, we conducted accuracy verification and evaluated the developed system. Finally, using the detection results created by using deep learning, we constructed a detection system to calculate the cost from the resulting image using the 3D model.
UAV is attracting attention because it can be used to grasp the situation during disasters, create maps, and maintain structures. Generally, when creating 3D models using UAV aerial images, SfM/MVS software is used. However, images taken with UAV contain defocus images due to the effects of device shaking. Therefore, we need to manually remove defocus images from a large number of images. We focused on the fact that UAV aerial images are continuous images with a high wrap rate. In this research, we proposed a general-purpose method for extracting defocus images using the difference in edge rate between the previous and next images.
In Japan, Dynamic maps are used for automatic driving development. Dynamic maps utilize point clouds obtained by MMS, and it is necessary to realize efficient data processing and labor saving. On the other hand, in the past research, the construction of algorithms for automatic extraction is a future subject. In this research, we have studied algorithms for automatically extracting road lane lines from point cloud data. We used the random forest method for a machine learning. In the optimal combination of feature quantity and learning data, accuracy of correct answer rate was 97.7%. The proposed method has high extraction performance even compared with existing methods.
The use of 3D point cloud data by MMS is attracting attention. However, the amount of point cloud data is enormous, and there is a need for a method to maintain data uniformly and efficiently. In this study, the authors constructed a method for detecting sidewalk barrier information using point cloud data, and examined the conditions that can be used for the development of walking space network data. As a result, the point cloud data obtained by high-performance MMS was able to grasp the features and topography at a point cloud density of 100 points per square meter or more. In addition, when the proposed method was examined from point cloud data for each point cloud density, barrier evaluation was possible when the point cloud density exceeded 1,000 points per square meter and the grid size was less than 0.10 m.
In a smart society, physical space and cyber space are integrated. Similarly, the construction project life cycle is being considered to be mutually related on both physical and cyber spaces. The project life cycle becomes complicate when integrating the relationship between physical and cyber space. This paper presents the design of a system architecture model in which the project life cycle is hierarchized, which is based on the scope of each process in the project. Additionally, an analysis of the information technology that supports each process from the scope of this model is also presented. The paper concludes that the proposed model is a useful tool to the construction project life cycle, and it helps to organize the process and system of the project components.
The recent development of the SfM technique is improving the efficiency of infrastructure management and quantitative evaluation of structural members using such technique is ongoing problems. In this paper, a dense point cloud model of a damaged T-shape steel structural member was constructed from still digital photographs. The 3D point cloud model was converted to an FE model by downsampling of the dense cloud in 3D lattice space and continuous 2D Delaunay triangulation of each section of the member. The reference FE model based on the measured dimension of the specimen was also composed. From comparisons among static loading test of the specimen, and FE analyses of both the point cloud FE model and the reference FE model. Fundamental availability of the FE modeling from SfM was confirmed that the distribution of stress in the web of the steel member consistent with the reference FE model. For the future study, the modeling accuracy of the thickness of the web plate in the point cloud FE model needs to be improved and the modeling algorithm is also modified to avoid defectiveness of solid elements.
Road administrators perform traffic censuses for road maintenance and management. In these censuses, surveyors visually count the number of passing vehicles, labor-saving method is necessary because the surveyors must count in many places and it is time-consuming. Mechanical surveys have been introduced as a countermeasure. However, considering the growth of ICT, further development in survey technology is expected. In the field of urban transportation, various technologies are suggested to recognize vehicles from video data using deep learning. For putting it to practical use and spreading some methods widely, it is necessary to survey the practicality of technologies and proper conditions of photographing based on the existing studies. In this research, we developed a technology of recognizing vehicles based on existing methods and took videos under various conditions. By analyzing them using the present technology, we clarified the usefulness of deep learning in the proper conditions of photographing.
In recent years, with the development of laser measurement technology, the means to measure 3D information of road features as point cloud data has been diversified, and the use of point cloud data has been expected. However, since point cloud data does not hold information about road features indicated by points, it is difficult to operate efficiently according to the application. Therefore, the technology to extract road features from point cloud data is attracting attention. In the existing research, a method to extract road features from point cloud data by referring to outline information described in the plan of completion drawing has been proposed. However, the point cloud data of MMS generally used in the road fields has a problem that the position is shiftting between the point cloud data for each trajectory or the point cloud data and the plan of completion drawing. In this research, we propose an improved accuracy method of extracting road features from point cloud data by registration of this positional shift.
In the construction site, the content of the safety instruction changes every day even in the same place. So far, on a system using beacons, there was a function capable of announcing determined contents when receiving a signal, but it was a problem that change of contents was difficult. The safety reminder tool developed this time can change the announcement content in real time in the network environment. By this, the safety instruction in the morning assembly can be confirmed by hearing the announcement of the safety information again on the site. Therefore, the three realism which can confirm the reality by seeing the actual thing on the site can be realized. And, since this system can deal with multiple languages including English, etc., the safety instruction can be given to the foreign workers who are expected to increase in the future by the native language, and it is considered to be connected with the safety improvement on the site. Since the multilingual correspondence can be utilized for on-site guidance, etc., this technology will be also utilized for the positive publicity on the site.
In 2014, a regular inspection of public infrastructure facilities (bridges, etc.) once every five years became a legal system. Of the approximately 730,000 bridges nationwide, 530,000 are small-scale bridges with a bridge length of 2-15m. These bridges are often not easily accessible by inspection engineers. Therefore, in this study, for the purpose of studying inspection methods as an alternative to visual inspection for small-scale bridges, development of inspection robots, examination of efficient methods for creating 3D CAD data, inspection using 3D models. We examined the results from four viewpoints, the management system of results and the extraction of damaged parts from photographs by AI. As a result of the research, we conducted demonstration experiments several times, arranged the scope of technology application according to the usage scene, and clarified the definition of functional requirements. This can be expected to contribute to the development of more efficient inspection, improvement of productivity, safety, etc. in future small-scale bridges, as a material for development for actual operation.
ICT construction machinery uses under-the-machine control / machine guidance systems at construction sites to obtain log data on their movement and behavior. These data are expected to be increasingly utilized and distributed. The authors have proved in an existing study that there is a possibility that three-dimensional shapes of objects of construction works can be generated from the log data of ICT construction machinery used in road pavement construction works. Findings were also obtained that as-built management using log data can be realized if a correction method capable of calculating the pavement edge of the construction boundary can be established.
In this study, a method for correcting log data was devised, and the existing method for generating three-dimensional shapes was improved. Then as-built management was performed as a trial in a test execution of road excavation and pavement construction, and it was proved that the improved method can be applied in practice.
In Japan, bridges need to maintain for connecting the road network and for preventing disaster. For visual inspection to bridges, it is important to access inspection data neseccaly because of lack of inspection expert engineer. And also, recently as ICT technique is advanced, AR(Augmented Reality) technique has been developed.
Normaly, when inspectors check the bridge visualy, they bring the bridge data as basic specification and historical inspection data and so on on printed on site. But hard printed data is a limit to the amount for carring. Furthermore, detailed data cannot be accessed in this system. If inspectors want to know it detail data, they go back to management office. It is not convenience and efficience.
In this study, it is aim to be efficiently at the visual inspection for bridges, firstly “Bridge-Card” such as name card size and drawing bridges on print is made, next when smart phone camera is taken the “Bridge-Card”, this system shows the basic specification, history of inspection and history of repair on smart phone. As results we were developping visual inspection support system. And it is cleary to improve how to the rate of identification at making “Bridge- Card”. This system improved as operation addtionala function to touch operation. Finaly, we proposed to use how to system on site.
Ministry of Land, Infrastructure, Transport and Tourism had amended the Road Law Enforcement Order at 2014, in order to find early damaged conditions and visual inspection is required every 5 years by the road administrators. Therefore, the amount of inspection work increases. Hence, it is need expert inspector for bridge maintenance who have specialized knowledge. With this background, many workshops are held in everywhere as a bridge inspector to maintain bridges. However, normally workshop is carried out on-site. That is to say, if weather condition is rainy, workshop is canceled. And another condition such as traffic and safety are problem to be held in workshop.
Therefore, we focused on MR technology that can realize an immersive experience in virtual space. In this study, we were developing a bridge inspection experience system with an MR Head-Mounted-Display that can simulate inspection work in a VR space for the education. The target of the bridge material is the inspection of the reinforced concrete bridges. 3D-VR models is made from real bridge photos. Thus, a young inspector can be experienced proximity visual inspection. In addition, in the impact inspection experience, the young inspector shakes the motion-controller instead as a checked hammer and hits the bridge, and marks damaged area using the hitting sound.
In this paper, we describe how to developed this system and how to utilize, and through the developed system as a result it show the efficiently to use MR technique for bridge inspector.
In urban areas, it is expected that the number of businesses such as reconstruction, renewal, and the construction of new structures will be increased in areas where there is a large volume of remarkably deteriorated infrastructure equipment and underground buried objects. However existing structures are usually managed by two-dimensional drawings, it is difficult to determine the position and height of underground objects during construction works.
The purpose of this research is to establish efficient correction system for the existing 2D drawings, by using a simple 3D model generation method for underground buried objects. In this research, the authors have developed an existing 2D drawing correction system, applied a 2D drawing correction method using a 3D model in the realization field, and verified the accuracy of a 3D model and the 2D drawing correction method. As a result, the authors extracted image data from the generated 3D model, and confirmed that the drawings can be corrected quickly by combining the extracted image data and the existing 2D drawings.
Due to heavy rain in northern Kyushu in July 2017 and the Eastern Hokkaido Iburi Earthquake in 2018, many landslide disasters occurred in a wide area. Satellite remote sensing technology is effective when it is difficult to identify disaster spots in a wide area or to approach a mountainous area immediately after a disaster. In this study, in order to extract landslide disaster areas based on differential images of high-resolution satellites before and after the disaster, a threshold correction method was considered to separate disaster areas and non-disaster areas taking into account the land cover of the affected areas. In addition, in order to more appropriately evaluate the extraction accuracy of disaster areas by satellite image processing, evaluation was performed based on reference data corrected by visual interpretation of satellite images after the disaster. As a result, the accuracy improvement by the threshold correction method here was shown. In addition, by using both GIS technology for pixel complementation of analysis results and slope information of the affected areas, erroneous extraction of landslide disaster areas was reduced.
If the information generated by routine work can be used for extraction of construction work whose conditions are designated, and the extraction can be done automatically, the efficiency will be significantly improved compared to creating a database and manual work. In this report, we picked up construction ordering documents as information to be used and showed the results and evaluation of experiments in which each page type is automatically classified into 8 classes by CNN (convolutional neural network). Through experiments, it was possible to extract a core page in tender notice or a covera page in implementation design documents from Shizuoka Prefecture's general civil engineering ordering documents and obtain an outline of the construction with 50 training data per class. The classification performance is improved by increasing the number of data, and it can be applied to the classification of construction ordering documents for road map updates. This method is expected to be applied to construction deliverables such as electronic ones.
The Ministry of Land, Infrastructure, Transport and Tourism published a manual of finished shape by TLS in 2016. Use of TLS is expected to increase in paving work. In the study measurements were made using TLS with the three different method. As a result, the time became longer with more measurements. In data processing, if there are many orientation points, the time required for coordinate conversion will increase. On the other hand, the noise processing time was shortened when the measurement range was narrow. Measurement accuracy increases as the distance between ground control points decreases. The error may be large near TLS installation.
MMS attracts attention as the system which can acquire information of feature around road with high accuracy and efficiency while driving. However, there are an extremely low number of lationship between the driving speed of MMS and measurement precision.In this study, we compared the precision of measurement results by the MMS systems when driving speed at 22km/h and 47km/h. As a result, RMS errors were smaller value at 22km/h run by 0.023m than at 47km/h run with the absolute precision. Also, RMS errors were smaller value at 22km/h run than at 47km/h run in comparison with each target.
Currently, satellite positioning technology is drawing attention with the aim of putting autonomous driving into practical use. Therefore, this study did verification of the combined use of QZSS in satellite positioning during street driving. As a result, it was shown that the fix rate was improved by 36% by using QZSS together in this case.From these results, the combined effect of QZSS was clearly recognized, but the fix rate was not high in any case. Therefore, analysis from the positioning results revealed that the increase in the number of satellites strongly contributed to the improvement of positioning accuracy.
Recently in Japan, traffic accidents caused by elderly drivers become a serious social problem. It is said that these traffic accidents are often caused by a decline in physical function with age. The realization of automated driving has the potential to solve this problem but it is difficult that the introduction on urban roads. Thus, it is important to develop technology which prevents before occurring traffic accident.
In this study, we propose a method to detect abnormal driving in an intersection because there are many traffic accidents caused by misoperation of an elderly driver in an intersection. A smartphone was set in the car and it obtained the vehicle characteristics such as acceleration and steering angular velocity. It was constructed the abnormal driving detection method applying One-Class Support Vector Machine (OCSVM) by using these data.
In Japan, utilization scene of three-dimensional data measured by instruments of laser scanners and cameras expands with i-Construction politics. So far mobile mapping system has been used in roads. However, it is not suitable to measure in the place where vehicle cannot enter or is difficult to use satellite positioning. For this reason, development of portable laser scanners is flourishing. The instruments are usable for substructure of bridge, mountainous regions, and indoors. But various sensors into the ready-made instruments have different installation angles and accuracy. And, the algorithms of data processing are not released. For these two reasons, it is difficult to analyze the factors of decreasing measuring accuracy in the ready-made portable laser scanners. In this research, an original portable laser scanner was proposed. We surveyed ready-made portable laser scanners and experimented to measure with prototype. And, we designed a portable laser scanner based on these characteristics and usage considerations for measuring three-dimensional data on structures.