SOME KEY TECHNIQUES ON UPDATING SPATIAL DATA INFRASTRUCTURE BY SATELLITE REMOTE SENSING IMAGERY

摘要

In general only a small part of spatial data in SDI change because of different factors, so rapid and effective updating methods are very vital. With the improvement of spatial resolution of satellite remote sensing (RS) imagery, it is possible to update spatial data infrastructure efficiently by satellite RS images. But the RS data volume is vast, so high processing capacity is required. It is impossible to update spatial data without the support of high performance of RS image management, retrieval, and change detection and pattern discovery. After introducing some background information including significance of SDI updating, feasibility and advantages of satellite RS imagery used to update spatial data, the framework of updating spatial and attribute information based on RS image is proposed. In the process, the integrated processing of raster data and vector data is very important, so some functions of RS image processing and GIS software should be fused. In order to extract the anticipated images from vast image database, effective retrieval technique is vital. Based on content-based image retrieval, content-based RS image retrieval is put forward and some topics including retrieval pattern, useful image content, feature extraction and similarity measure are researched. After related images are retrieved from vast image database, it is necessary to discover those change areas, so change detection from multi-temporal RS images are discussed further. RS image data mining and knowledge discovery is the synergy of Spatial Data Mining (SDM) and Image Data Mining (IDM). Oriented to the demands of SDI updating to intelligent information processing, some primary issues on RSDM are analysed. It is pointed out that updating SDI by satellite RS imagery will be potential advantageous in the future, and all the techniques discussed in this paper including content-based RS image retrieval, change detection, multi-temporal RS image fusion and RS Data Mining and Knowledge Discovery are very important and will play important roles in the future.