nt 
re 
nt 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
Where Nearest (Snapshot (position, NOW), Point (x, v), k) = 
TRUE F 
(2) Slice Queries 
In this section, we show the example of the slice queries for 
moving objects. The slice query requires the period parameter 
as follows. 
Example 6: Find the location of cellular phone user 1001 
between time // and #2. 
Select Project (Slice (location, 17, /2)) 
From People 
Where id = 1001; 
Example 7: Find people who were within Ikm from point (x, y) 
between time // and /2. 
Select id 
From People 
Where Withins (Slice (location, 11, 12), Buffer (MPoint (t1, t2, x, 
y), 1000)) = TRUF; 
Example 8: Find k people who were closest to the point (x, y) 
time // and 2. 
Select id 
From People 
Where Nearest (Slice (location, #1, 12), Point (x, y), 1) = TRUE; 
(3) Trajectory Query 
Trajectory query include the operations such as Enter(), Leave(), 
Passes(), Insides(), Meets(). 
Example 9: Find people who were (at least once) in Daejeon 
between time // and /2. 
Select id 
From People, Region 
Where Passes (Slice (location, 17. 12), Region.area ) = TRUE 
AND Region.name = 'Daejeon'; 
4. BUFFER MANAGEMENT MODULE (BMM) 
The BMM plays an important role in enhancing the 
performance of the query of location insertion. Locations of 
moving objects are permanently stored into various types of 
database systems by LSM. In this environment, it is difficult for 
LSM to process every insertion request from QPM, because the 
cost of insertion transaction is very high in database system and 
insertion requests from QPM occurs very frequently. Therefore, 
buffering of insertion request and batch processing are very 
effective. 
Another role of BMM is that when QPM issues a search request 
BMM searches moving objects in the memory buffer, transfer: 
the request into LSM, and then bind the results from the 
memory buffer and those from permanent storage. 
4.1 MORow Object 
BMM stores locations of a moving object into a MORow object 
for insertion request from QPM. There is one-to-one 
relationship between a moving object and a MORow object. 
Therefore, a MORow object take a responsibility for buffering 
locations of a corresponding moving object. MORow object 
depicted in Figure 8 is composed of MOID (Moving Object 
113 
[Dentifier), Length (the number of locations stored in it), MBR 
(Minimum Bounding Rectangle of the locations), From (time 
that first location in it is acquired) and To (time that last 
location in it is acquired). The MaxLocation means the 
maximum number of locations stored in a MORow object. If 
the Length of a MORow Object is equal to the MaxLocation, 
then all of the locations in the MORow object are transferred to 
LSM in order to store permanently. 
  
Memory Buffer 
Pa "AGI 7 o00MM 
ROID 
  
  
  
  
  
  
  
  
  
ect ne ne acc tram ren pére ta ne net sin netteté met 
? 
i MOR ew 
| E 307 3E NI RENN 
— 
  
  
Leesen[X, Y, Bere. Ties) 
Lecerse[X, Y, Error, Timds 
[eren T eren, Ti 
[1 [ [LecatinfX ¥. Brror, Time) 
i 
  
  
  
Figure 8. Structure of Memory Buffer 
4.2 Memory Buffer 
The overall structure of memory buffer is composed of a set of 
MORow objects. Each MORow object represents a trajectory of 
a moving object from From time to To time. As the figure 
indicates, there is a B-tree for indexing MOIDs of MORow 
objects. The BMM, therefore, finds efficiently a corresponding 
MORow object by using MOID. 
5. LOCATION INDEXING MODULE (LIM) 
According to the previous works (Pfoser et al., 2000; Kollios et 
al, 1999; Nascimento and Silva, 1998; Vazirgiannis et al., 
1998; Song and Roussopoulos, 1987), there are three kinds of 
location indexes for moving objects. 
Current Location Indexes: The indexes of this type take only 
current locations of continuously moving objects into 
consideration. And current locations are also used for 
anticipating future locations of moving objects. These indexes 
should have capabilities to process frequently updates of 
numerous moving objects. 
Past Location Indexes for time interval (or time point) queries: 
The indexes, such as 3DR-tree and HR-tree, have a special 
purpose of efficient processing of a time interval (or time point) 
queries for the current and past locations. 
Past Location Indexes for trajectory queries: The indexes, such 
as STR-tree and TB-tree, have a special purpose of efficient 
processing of a trajectory queries for the past locations. 
LIM we implemented in this paper supports all kinds of indexes 
mentioned above. We implemented Adaptive Quad-tree (show 
Figure 9) as current location index. This index partitioned