| Summary: | 碩士 === 國立海洋大學 === 電機工程學系 === 89 === The integrated application of Internet and digital mobile information is future trend. Database plays an important role for Internet provider. How to fast search a digital database becomes more and more necessary. Satellite image application becomes available to all. At present day, it is essential to develop an efficient indexing technique for database retrieval. In this thesis we propose an image indexing technique for fast image retrieval and by means of this we build a multi-spectral satellite image database.
Typical satellite image set exhibits a number of different terrains with different spectral characteristics. In order to achieve fast retrieval for multi-spectral satellite image, it is necessary to segment the image and label cover feature. In this thesis we first propose an Eigen Region Indexing technique which utilize KLT and extracted principle eigen vector. We adopt the principle eigen vector as our image index.
Then we build up a learning mechanism for multi-spectral satellite image database which combines the Eigen Region Indexing technique with SQL. By means of the extracted image feature, we can retrieve image with similar feature. Supervisor can train indexing labels for user retrieve and users can train for supervisor reference. By this way, we build a interactive satellite image database retrieval system. The simulation results show that this technique is suitable for multi-spectral image feature identify.
|
|---|
