| Summary: | 碩士 === 國立臺灣大學 === 資訊管理研究所 === 90 === In the coming years, the population using wireless devices will grow rapidly. With providing special services for each individual customer, the wireless devices are becoming more and more popular. Tracking the positions of moving objects is hard for the indexing methods of the current database systems. The reason is that most of proposed indexing methods are suitable for still(not-moving) objects, not for moving objects. Since the positions of moving objects are changed over time, it’s impossible for the indices in the database to keep the correct positions of moving objects. If we try to update the positions of moving objects immediately when they move, the cost is really high.
In this thesis, we first propose the MBR-tree with the minimum boundary velocity for each bounding rectangle. This makes the bounding rectangles not grow so rapid and the overlapping areas are reduced effectively. The MBR-tree uses a bounding rectangle to cover all the objects or rectangles within the bounding rectangle. This motivates us to use a different shape to enclose moving objects. So, we propose the MBS-tree that uses a bounding sphere to enclose moving objects. We also propose an algorithm to compute the minimum boundary velocity for a bounding sphere. The MBR-tree and the MBS-tree use a rectangle or sphere to enclose all moving objects. These methods will generate a lot of overlapping areas and such overlapping areas will reduce the performance of the MBR-tree and the MBS-tree. Therefore, we propose the MQ-tree in which the whole data space is partitioned into no overlapping regions. Next we classify all the moving objects into three types and 24 cases of movements. Then we recursively construct the MQ-tree according to the cases of movements.
|
|---|
