Augmented Reality is a technology for overlaying virtual objects in real space and is used in various fields. Markerless AR is attracting particular attention because it can be used indoors and outdoors without the need for markers. However, there are technical issues such as limited fields, registration, and real-time performance. Therefore, we propose a method of automatically creating markers from the environment map created by SLAM technology. We realize markerless AR by searching the created markers from the geometric data of GIS. As a result of the experiment, it was found that AR expression can be realized with high accuracy and in real time.

 Although disaster processing execution plans must be prepared as soon as the disaster occurs, it usually takes much time to draw them up. Thus, our research group has been developing an automated preparation support system for drawing up viable processing execution plans applying Artificial Intelligence (AI). In this research, an ontogoly was developed to represent the concepts, attributes, relationships of terms appeared in disaster processing execution plans and other related documents in order to systematize the functional body of knowledge of the vocabulary, concents, and semantics. Furthermore, a semantic checking system was developed, based on the developed ontology, for checking the template for disaster waste processing plan and processing execution plans and previous planning documents. the semantic checking system was verified by a number of actual cases.

 When transporting designated waste generated by the Great East Japan Earthquake, it is necessary to develop an optimal transportation vehicle operation plan that satisfies various constraints, but due to the increase in calculation time caused by combinatorial explosions, only approximate solutions have been obtained. In this study, we focused on the necessary constraints for making a transportation plan of designated waste, such as completion of transportation within working hours and prevention of overlapping arrival times of transportation vehicles, and made a basic formulation for deriving an optimal transportation plan by quantum annealing. We also formulated a cost function that can be used to recreate the operation plan when the plan is disturbed by delays that may occur during the execution of the transportation plan, and optimized the transportation plan assuming normal conditions and when delays occur. As a result of the optimization, it was confirmed that by setting appropriate parameter values, an optimal operation plan table could be output under various conditions.

 Centimeter Level Augmentation Service (CLAS) is characterized by its ease of use and low cost. Therefore, it can be said that CLAS is suitable for use in vehicles. In this study, we conducted kinematic positioning experiments using CLAS (AQLOC-Light) and RTK positioning, compared the location data of both, and verified the accuracy of CLAS. As a result of the verification, the Fix rate for each experimental route tended to be higher for CLAS than for RTK positioning. In addition, we confirmed that CLAS tends to hold a higher Fix rate than RTK positioning even when the number of observation satellites decreases. Further-more, in the evaluation using the Fix solution of RTK positioning as the true location, it was confirmed that few CLAS positioning solutions have a large horizontal error and that the Float solution is highly stable. From these analysis results, the superiority of CLAS in kinematic positioning was shown.

 Since the aging of a large amount of infrastructure built during the high growth period has become a problem, the maintenance, and management of ICT (Information and Communication Technology)-utilized infrastructure is an urgent matter in Japan. Although UAV is an up-and-coming tool in 3D data generation for existing structures, the operation method for complex-shaped structures is ad hoc. SfM (Structure from Motion) of the photogrammetry process is a compelling 3D digitization method. However, the processing cost is very high, primarily when aerial photography with UAV handles many photographs. Also, the rework caused by imperfections in aerial photography is a substantial loss. This paper proposes a method to monitor whether aerial photos are suitable for SfM by visual SLAM (simultaneous location and mapping) process in real-time to prevent such rework. Visual SLAM, which also uses the photogrammetry principle, can track the camera's egomotion from the captured image sequences. By taking aerial photographs at various speeds in advance and investigating the causal relationship with the quality of SfM, the appropriate limit of moving speed is determined. Using this speed limit as an index, SLAM monitors the moving speed in real-time to evaluate aerial photography's appropriateness. We also validated the derived index by experiments based on the implemented proposed method on an actual UAV.

 In recent years, advances in satellite positioning technology and the widespread use and sophistication of mobile information terminals such as smartphones are expected to lead to the realization of a society with a high-precision positioning environment (high-precision positioning society). Narabe et al. conducted a 3D point cloud measurement experiment using an indoor MMS, and created a 3D model for indoor navigation. However, the simplification of point cloud processing and modeling work remained an issue. Therefore, in this study, assuming that the 3D model of indoor environment is generated from the point cloud acquired using a walking MMS, we evaluated and discussed the automatic removal of noise in the acquired point cloud and the method to simplify the work of generating 3D data from point cloud data.

 The authors have developed a Cable-stayed bridge cable inspection robot to improve the efficiency of cable inspection of Cable-stayed bridges. The six cameras are mounted in the robot and takes a video of the whole circumference of the cable surface at a single elevation. It is characterized by making an expanded view of image from the taken video. In this paper, the red LEDs on the green circuit boards are installed in the Cable-stayed bridge cable inspection robot in order to synchronize the cameras. Then, a time synchronization method using image processing is proposed to automatically detect the red LEDs from the taken video. In addition, application to the tube which imitated the cable is presented so as to demonstrate the accuracy of the automatic detection method.

 The whole country road street traffic situation investigation set by the Ministry of Land, Infrastructure and Transport is having for its object to get the basic material to grasp the present state of the whole country road traffic and a problem and settle on a maintenance plan on the road for the future¹⁾. The investigation method has a road condition investigation by road register and measurement and a overall travel speed investigation by traffic count by a mechanical observation, ETC2.0 probe information and the public car probe data. However, a car with ETC2.0 is about 6 % of the whole country vehicle possession number (March, 2020). The area which can be acquired is decided, and ETC2.0 probe data can't acquire position information of all networks. Therefore, I think the present road investigation isn't enough for volume of information to grasp the traffic situation, and position information of all vehicles and position information of all networks are more necessary than the abovementioned thing. In this study, a video investigation was put into effect in a road just before the signalized intersection and the vehicle location of the driving lane and the passing lane was acquired from video data. And the vehicle trajectory was made from the vehicle location. Further, time headway was calculated using the vehicle trajectory data and car following model. I have for my object to grasp, compare and consider the congestion situation of 2 lanes of the driving lane and the passing lane.

 This paper presents MR visualization systems based on HoloLens 2 to support the design, construction and maintenance of underground structures. We developed two MR visualization systems: 1) MR visualization system for blueprints and 3D CAD models to support the planning and design stages, 2) MR visualization system for construction sites to support the construction and maintenance stages. To investigate the validity and effectiveness of these systems, we applied them to underground structures at construction sites.

 Currently, in many cases of pedestrian traffic measurement, pedestrians are counted by field surveying or a visual check of motion images. Accordingly, there are many problems including error counts caused by human error and the risk of heatstroke caused by outdoor measurement for long hours. In addition, in the environment where unspecified large numbers of people come and go, a problem of a decrease in pedestrian measurement precision due to occlusion has been revealed, and there is no solution established for this problem yet.

 In this study, a practical method for pedestrian traffic measurement was derived by verifying the technology of person recognition using deep learning capable of treating occlusion, and clarifying its problems. As a result, it was found that even if most of the person region is hidden, the number of pedestrians can be measured with high precision by applying the proposed method.

 In recent construction work sites, a large amount of point cloud data for various places have been measured and accumulated along with the technological innovation of survey instruments such as mobile mapping systems (MMS) and ground-based laser scanners and cameras. The point cloud data are used mainly for preparing maps and recognizing the shapes of structures. Nevertheless, since the point cloud data are an aggregation of massive points, if the semantic information of road features, such as dividing lines and traffic signs, can be provided with the data and recognized by the computer, increased efficiency and advancement of work as well as versatility can be expected.

 In this study, location information of road features, such as dividing lines and traffic signs from a dynamic map, which are high-precision data, is analyzed, and a method of converting it into area data conforming to the “Attribute management specifications for point cloud data [for roads] (draft)” is devised. Then, with a case study conforming to the devised method, it is clarified that the point cloud data of road features can be extracted using the area data.

 This paper presents a fast and accurate classification method for underwater objects using underwater mapping data obtained by a small Autonomous Underwater Vehicle (AUV). For the mapping data, in addition to underwater acoustic reflection intensity images, water depth data and backscattering reflection intensity data are employed. We proposed the classification method based on multimodal deep learning using a convolutional neural network. In order to verify the effectiveness of the present method, we applied it to the measured several underwater mapping data.

 In Japan, natural disasters such as typhoons, earthquakes, and tsunamis arise frequently, and the danger of block walls collapsing when a disaster occurs has been noted. Therefore, there is an urgent need to understand the current status of block walls and inspect them. Against this background, researchers of Tokushima University has been conducting field surveys based on the Ministry of Land, Infrastructure, Transport and Tourism's inspection index with the cooperation of the prefectural government and municipalities. However, block walls are scattered all across the country, and the current inspection method, which dispatches workers directly to the site, cannot cover a wide area. Therefore, we developed a method for extracting block walls from the point cloud data measured over a wide area and have been working on labor-saving of the field survey. In this research, we propose a new method for the risk judgement of block wall based on height and inclination, using the point cloud data.

 In Japan, taking the opportunity of the proposal of Society 5.0, the introduction of state-of-the-art technologies, such as IoT and AI, has been being examined. In particular, at construction sites, such technologies are expected to greatly contribute to an improvement in the productivity and development of safety management. A development of technology for managing the positions and conditions of workers to decrease accidents of minor collisions or falls has been attracting growing interest. Thus, paying attention to the safety management of construction sites, the authors have proposed the some methods of person identification by deep learning focusing on patterns pasted on the workers’ helmets. In the existing researches, however, there was a problem that the same person was not correctly identified successively between frames when erroneous identification occurs due to the difference in the distance from the camera or the way the pattern was reflected. In this research, a new correction method is devised based on tracking of the helmets and improved the person identification method. As a result of conducting demonstration experiments, knowledge was obtained that this correction method is effective.

 In Japan, mobile mapping systems (MMSs) are used for the maintenance of civil infrastructures, such as roads and rivers, and generation of high precision three-dimensional maps for autonomous driving. Using MMS is difficult for local governments because it is very expensive. Therefore, we developed a vehicle-mounted sensing unit with commercially available and inexpensive sensors, and we evaluated its practicality. We considered the self-position estimation of the sensing unit using SLAM (Simultaneous Localization And Mapping) because GNSS (Global Navigation Satellite System) could not provide a fixed solution in environments with viaducts or skyscrapers on both sides. However, when the measurement data includes vegetation, the estimation accuracy decreases. In this research, the error factors of self-position estimation accuracy are investigated and a method for correcting the errors by removing their factors is proposed. In the result, the method was verified in some experiments, we confirmed it was useful.

 In Japan, it is expected to conduct exhaustive maintenance of social infrastructures for roads and rivers since natural disasters, such as typhoons and earthquakes, occur frequently. In the field of roads, it is necessary to prevent the interruption of evacuation routes caused by the collapse of structures generated from road earthworks right after the occurrence of a disaster.

 For this reason, the implementation of specific earthwork inspection for specific road earthwork structures, including high fills and long cut slopes, has been made compulsory. However, it is difficult to manage and grasp the correct locations and detailed present conditions of all the road slopes. In this research, using the point cloud data acquired by MMS or airborne laser surveys and the road alignments, a method for automatically extracting the point cloud data of road slopes is proposed. Then, verification experiments were performed to validate the applicability of the proposed method to the actual construction sites.

 The Ministry of Land, Infrastructure, Transport and Tourism is promoting “i-Construction” to improve the productivity of construction sites by utilizing 3D information. Most of the 3D information showing the current topography measured by the laser scanner is stored as point cloud data. The point cloud data generally holds only 3D coordinate values and reflection intensity, but RGB values can be added by using camera images, and it is widely used due to a high visibility of features. However, there are some issues that require labor to select camera images corresponding to each point, and that cannot be colored by the equipment without a camera. Therefore, in this research, we develop an automatic coloring technology using GAN for point cloud data that does not require a camera image.

 In Japan, road administrators perform traffic censuses to understand the status of automobile traffic. In this census, it is common for investigators to visually check the automobiles and count the numbers of passing automobiles for both small and large cars. But, it is difficult to secure workers because the working age population is decreasing. Therefore, the Ministry of Land, Infrastructure, Transport and Tourism will abolish the manual survey and consider introducing some techniques that automatically count the number of passing automobiles using video images. In existing research, techniques for the classification of automobile type have been developed using machine learning or deep learning. These techniques are not yet accurate enough to be used in practice. In this research, we develop a technique to count the number of passing automobiles for each automobile type. This technique classifies the automobile type based on the outer shape of the automobile and the shape of the parts, which investigaters pay attention to when they classify the automobile type manually. Furthermore, we clarified the usefulness of the proposed technique through some demonstration experiments.

 In Japan, road administrators perform traffic censuses to understand the status of automobile traffic. Recently, in these censuses, techniques for counting the traffic volume from video images have attracted attention for the purpose of improving work efficiency and to save labor. These techniques can count with practical accuracy in video images taken in the daytime. However, the counting accuracy is reduced in video images taken in the nighttime as sufficient brightness are not secured and a shape and color of the vehicle are obscured. In this research, we develop a traffic census technique for application to nighttime traffic using existing techniques. This technique converts video images shooted the nighttime into video images taken in the daytime using deep learning. Furthermore, we clarified the usefulness of the proposed technique through a demonstration experiment.

 In recent years, effective safety management measures based on the Internet of Things (IoT) have been established to prevent industrial accidents in construction sites. In response to these efforts, technologies that use object detection methods to obtain the location information of workers and construction vehicles can contribute toward safety management. However, it is difficult to detect workers and construction vechiles from video images by object detection methods in construction site where dangerous areas are constantly changing due to their coexistence. Therefore, the application of deep learning can be considered, but in order to detect them accurately, it is necessary to update the detection model specialized for construction sites. However, it takes a lot of cost to manually create a training data for constructing model. Therefore, in this research, we propose an automatic generation method of training data for deep learning to realize high-precision for detecting workers and construction vehicles in construction sites. Experiment was performed with application of the proposed system to the video data at construction site, and its usefulness was confirmed.

 In Japan, owing to the steadily declining working population, innovative technologies that improve productivity in construction sites using i-Construction are urgently required. This necessitates the tracking of not only construction vehicles but also workers using technologies such as AI and IoT. Against this background, many technologies that were able to track both workers and construction equipment on construction sites had been developed. However, challenging subjects for automating tracking tasks remained as those technologies were applied in sections where extended and heavy occlusion occurs frequently. Therefore, an identification method that matches appearance features of workers and construction vehicles before and after occlusion is proposed in this research, thereby contributing to the development of tracking technologies that enable successive tracking through occlusion sections. Experiments conducted using videos from actual construction sites showed that the method was capable of restoring tracking progress in occlusion sections by identifying each worker and construction vehicles individually before and after each section.

 In order to consider effective accident countermeasures for road safety projects in cooperation with road managers and traffic managers, it is necessary to develop a simple accident database system incorporating various information. In this study, we proposed a new analysis system with geographic information and other data to build a database that can be easily used by a wide range of people. In addition, we studied a simpler map matching method to combine the route information and the accident data. We have conducted a questionnaire survey on the usability of this database, and confirmed the consistency with the needs of users. Using the results obtained from this database, it is possible to understand the distribution of occurrence of specific accident attributes and to visualize the relationship with the existing countermeasures.

 Quasi-Zenith Satellite System (QZSS) is a Global Navigation Satellite System (GNSS) that is currently developing for improving the accuracy of satellite positioning in Japan. This study verified the positioning accuracy based on the combined use of QZSS and other GNSSs. In addition, the positioning results obtained by comparing different combinations of satellite positioning systems were evaluated to determine which combination of satellite positioning systems provides the most efficient and accurate positioning. As a result, under an open skies environment, the combination of the QZSS and GPS can provide stable and high positioning accuracy, but under a radio-shielded environment, it was found that the adding BeiDou or Galileo and using three GNSSs can improve positioning accuracy.

 Positioning technology using a positioning satellite system is required for the development of next-generation ITS technologies such as autonomous driving. The number of positioning satellites that can be captured is small in an environment where high-rise buildings are lined up or the sky is covered with vegetation and signals affected by multipath such as reflected waves and diffracted waves, greatly reduce the positioning accuracy of satellite positioning systems. In this study, the difference in signal strength was obtained for each vehicle behavior whether the vehicle was running or stopped, and the effect of selecting a satellite suitable for positioning calculation was verified. As a result, it was confirmed that the positioning rate and the fix rate were improved depending on the driving direction, driving environment, and vehicle behavior.

 In this study, we analyzed and verified the acquired point cloud when the material of the target surface was different during measurement by MMS equipped with a phase difference laser scanner. As a result, the point clouds were plotted in the place different from the targets installation position for targets with strong reflection intensity, and noise was likely to occur and the shape of the point cloud was disturbed and the number of acquired points was reduced in the target with weak reflection intensity.