Major Research Subjects of Shi, Z. (1998-) Visiting Associate Professor

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

  1. Wavelet Based Image analysis
    Wavelet analysis is widely used for signal and image analysis in a lot of scientific domains because of its time-frequency localization property. The purpose of this research is to extract "useful" geo-features like building polygons from images. Edges were extracted from images with wavelet transform. Edges and corresponding direction information were then employed for image segmentation.

  2. 3D Building Extraction for Digital City Construction
    This research aims to extract 3D building from stereo photograph images. The key point of this research is polygon-matching algorithm. In this research, images are firstly segmented into polygons. Polygon-based stereo matching is then performed for 3D information acquisition, which is utilized to differentiate roofs from polygons on ground. The possibility and effectiveness of proposed methods have been evaluated with several experiments.

  3. Object Extraction from High Resolution Satellite Images
    1m-level High Resolution Satellite Images (HRSI) are available for private use now. People are interested in where these images can be applied. For example, can these images be used for urban information system construction or updating. In this research, the properties of HRSI are analyzed firstly. The algorithms related to building and road extraction were proposed. The possibilities related to building and road extraction under different conditions (e.g. building size and density, road type and wide etc.) are analyzed and discussed.

  4. GIS Database Revision
    The importance of GIS database has been recognized for a long time. But, it is still very difficult and time-taking a work to capture the spatial and non-spatial changes. This research focuses on detection of spatial changes by comparing the existing GIS database and image data. Two methods were proposed. One is vector-info based revision method, while the other is raster-info based method. Experiments show that the combination of two methods can highly increase the efficiency of geo-change detection.