Ryosuke Shibasaki

Professor, Division of Spatial Information Engineering

Further information is available at http://shiba.iis.u-tokyo.ac.jp/en/index.html

 
His research interests cover spatial data acquisition techniques of a real world, data model development and their applications especially in environmental fields in conjunction with agent-based models. Major targets of data acquisition and modeling are such fundamental data that can support diversified data uses as social “infrastructure” or spatial data infrastructure (SDI). The research is also extending to design or planning methodologies to determine what kinds or types of spatial data with what level of quality should be shared and developed as spatial data infrastructure (SDI). Target scales of spatial data are two-folds; street-block to city scale and continental to global scale.
Major research topics are summarized as follows;

 

Development of sensor systems and algorithms for automated spatial data acquisition

1) Automated acquisition of 3D spatial data at street-block to city scale
Sensor systems and data processing methods are developed to automate data acquisition and model-building of 3D spatial data of urban features (geographical objects) at street block to city level. Sensor system development focuses on how to integrate what kinds of different sensors such as laser scanners and high resolution linear CCD sensors, while the development of data processing methods targets data fusion of those different sensors for automated feature (geographical objects) extraction and modeling. Examples of the sensor systems under development include air-borne Three Line Sensor (TLS) and vehicle-borne laser mapping system (VLMS). Only one system exists similar to the TLS, which is developed by DLR independently. For VLMS, there are no equivalent systems. Data fusion methods for those sensor systems include registration/merging of image data acquired independently and extraction and reconstruction of 3D features such as buildings, roads, automobiles and trees from those image data.

2) Development of sensor systems and data processing methods for observing moving objects such as pedestrians and automobiles.
Sensor systems for moving objects include laser scanners and positioning devices consisting of wearable sensors such as accelerometer gyro and barometer, which enables to measure the trajectory and behavioral patterns of not only individual persons or automobiles but also their group. In addition, possibility of Pseudolite (Pseudo satellite), a kind of transceiver which emits signal similar to GPS’s, is explored through field experiment with urban environment to achieve “seamless” positioning from indoors to outdoors. By combining Pseudolite with the laser scanners and the wearable positioning devices mentioned above, human behavior can be monitored in 3D urban space.

3) Automated spatial data development using satellite remote sensing at continental to global scale.
Algorithms of handling satellite remote sensing images are developed to generate spatial datasets such as DEM (digital elevation model), land cover dataset, temperature/precipitation datasets at continental to global scale.

 

Design methodologies for spatial data infrastructure (SDI)

“Spatial data infrastructure (SDI)” should be designed as such that can support the development of as many applications as possible as a commonly shared dataset. Selection/definition of features, design of their scheme and evaluation of cost-effectiveness of SDI development should be conducted based on systematic survey and analysis of information usage of diversified user groups. A design supporting methodology is being developed including surveying methods of information usage, an organizing method to extract features and their logical structures referred by diversified user groups. This methodology is actually used to design a common road data model to support national road administration and to extract 3D road and underground/indoor space data for supporting ITS service for pedestrians.

 

Reconstruction of spatio-temporal changes of real world phenomena

Many real world phenomena such as traffic congestion and atmospheric conditions change spatially and dynamically. However, it is usually very difficult to observe those changes continuously in both space and time. It creates necessity to reconstruct spatio-temporal changes from fragmentary observational data. Reconstruction accuracy can be improved by incorporating knowledge or models on behavior or dynamics of object-phenomena such as aero-dynamic equations for atmospheric conditions. In addition to the development of a conceptual model representing dynamic spatial data including observation and event data with uncertainties, reconstruction methods based on GA (Genetic Algorithm) is developed. The method is applied to reconstruct a very long-term land use/land cover change of the Earth and human movement in urban areas.

 

Development of spatial agent models

As detailed spatial information of diversified objects including moving objects in 3D urban scene can be acquired, a foundation can be formed to model behavior of individual persons and automobiles in a space as agents. It is suggested that the dynamics of traffic congestion and movement of crowds can be more faithfully represented by agent-based models(ABM). ABM (AGENT-LUC) is developed to model land use changes in Thailand and Laos focusing on agricultural land use changes due to population growth and deforestation by slash-and-burn agriculture. The model also contributes to identifying advantages and limitations of the approach and research challenges in linking with spatial databases and in implement using parallel processing. The AGENT-LUC also represents crops as an agent growing based on agricultural operations such as water management and fertilizer applications.