X-B4, 2012 
  
oud latitude = 
s a buffer of 0.1° 
en line is a buffer 
rraphy data type 
  
  
ange) and 1000m 
ion of Brazil 
>sql extensions. In 
xgrading easier. 
se a database tem- 
ise to be spatially 
aking it harder to 
| access. With ex- 
TE EXTENSION 
onnection, and ev- 
lefinitions for each 
few lines defining 
from about 13,000 
iso more portable 
version backup to 
he new definitions 
conflicting method 
“ALTER EXTEN- 
2.9.1.2 of [5]). If 
l 
  
International Archives of the Photogrammetry, Remote Sensin 
     
g and Spatial Information Sciences, Volume XXXIX-B4, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
25 Support for images 
The new raster support permits processing images by using SQL 
queries. The images can be stored inside the database or “out- 
db”. This has not been tried in Inteligeo yet but it may be a good 
tool to provide users with only the image part they need. This is 
important in regions of poor network connectivity. 
3 RESULTS 
Inteligeo production database was migrated to a development ver- 
sion of PostGIS 2.0 in October 2011 and after 6 months of con- 
tinuous operation and very few updates, there have been no in- 
cidents with the database at all. Our database access is almost 
entirely composed of reads, so it may constitute a different envi- 
ronment from other deployments. 
Invalid geometries of new data are no longer being dropped. Ev- 
ery invalid feature is processed by the ST. Makevalid function. 
[ter p 
(a) Geometry before processing (b) Geometry after pro- 
cessing 
Figure 5: Spike removal of data before insertion into the database. 
Figure 5 is an example of spike removal, one type of process- 
ing used in Inteligeo. This procedure relies on buffer operations 
and can be performed safely only on valid geometries. Figure 5 
shows typical artifacts remaining from operations such as the in- 
exact subtraction of one geometry from the other. This is similar 
to the image processing morphological operations of an Closing 
followed by and Opening using radius 0.00001 [1]. The radius 
of the operations is adjusted according to the dataset, and the 
join=mitre’ option is used to preserve as much as possible the 
original geometry and not round the corners: 
ST Buffer( 
ST Buffer( 
ST Buffer(ST MakeValid (geom), 
0.00001, ’join=mitre’), 
jo: 00007, ’join=mitre’ 
0.00001, ’join=mitre’ 
) 
31 Application compatibility 
Postgis 2.0 supports the same queries as version 1.5.3. Our maps 
am Served using ArcGIS 10.0 connected to the database using 
query layers" (no ArcSDE is involved). Other than the labor of 
manually adjusting about 800 layer pointers (ArcGIS 10.0 lacks 
automatic adjustment of query layer sources), the new database 
Worked well, with no problems. We do most of our quality con- 
trol using QuantumGis, which also worked fine with the new 
database, except for Ticket #1200, which the team submitted and 
the problem got fixed in less than a week. 
3.2 Support 
The team reported three bugs on the website of the project during 
this time (Tickets 1197, 1198 and 1200). Two of them were ad- 
dressed on the same week (1197 and 1200) and I got a personal 
contact about the third one (1198) telling me that they would wait 
for more data, since the bug was very difficult to reproduce and 
may have been due to the GEOS library (not PostGIS). 
The community support for this project is very good. If profes- 
sional support is needed, it is possible to hire the companies that 
employ some of the PostGIS developers, with the added advan- 
tage that they may commit changes to the project code. 
3.3 Maintenance 
The extension support makes restoring from backup much easier. 
The main maintenance burden is tracking the source code releases 
of PostGIS, GEOS and GDAL and recompiling the source code 
every time an important bug is fixed in any of them. This has 
happened about 3 times in 6 months. Other than monitoring the 
releases, there are no significant tasks beyond regular database 
maintenance - keeping up-to-date backups, monitoring statistics, 
tuning and making upgrades between minor versions. 
ACKNOWLEDGEMENTS 
The authors would like to thank the Japan International Coopera- 
tion Agency (JICA, www. jica.go. jp/english, Alos4Amazon 
project), the UN. Office on Drug and Crime (UNODC, www. 
unodc. org) and the national Financier of Projects and Studies 
(FINEP, www. finep. gov. br) for sponsoring the project in which 
this work is inserted. 
References 
Gonzalez, R. and Woods, R., 1992. Digital image processing. 
Addison-Wesley world student series, Addison-Wesley. 
Group, P. G. D., 2012. PostgreSQL Featured Users. PostgreSQL 
Global Development Group. http://www. postgresql.org/ 
about/users/ (10 Apr. 2012). 
Multiple, 1991. The GNU General Public License (GPL 2.0). 
Free Software Foundation, Inc. http:/ /www.opensource. 
org/licenses/gp1-2.0.php (10 Apr. 2012). 
Multiple, 1996. The PostgreSQL License. PostgreSQL Global 
Development Group and The Regents of the University of Cal- 
ifornia.  http://www.postgresql.org/about/licence/ 
(10 Apr. 2012). 
Multiple, 2012a. PostGIS 2.0.0 Manual. Refrac- 
tions Research, Inc. http://postgis.refractions.net/ 
documentation/manual-2.0/reference.html (10 Apr. 
2012). 
Multiple, 2012b. PostgreSQL 9.1.3 Documentation. Post- 
greSQL Global Development Group. http://www. 
postgresql.org/docs/9.1/static/intro-whatis. 
html (10 Apr. 2012). 
Revised April 2012