Geo-Referencing Brazilian Public Telephones
Mapping
Brazil has a population of about 150 million and covers approximately 8.5 million square meters
of land (larger than the USA not including Alaska). The first problem the TELCOS had to face
was the identification of the localities. Localities were defined to be as little as an agglomeration
of buildings with some stable inhabitants. In certain regions, the identification of small localities
is quite simple due to the availability of precise maps. But there are also huge geographic areas
with dense vegetation, and even if good maps were available, the identification of small localities
would be quite difficult. Reasonable maps were available showing the municipalities (equivalent
to the US county) boundaries, roads and highways, and states boundaries. Other maps showed
specific locality boundaries that had to be adjusted to properly fit into the overall concession
map.
Ideally, meter resolution maps would have been used, but this was not feasible for the entire
country. Landsat images were used for almost every purpose: to identify urban footprints, to georeference
and rubber-sheet locality maps with problems, to serve as background for boundaries
and objects geo-referenced by several means, and so on.
To properly manage the locations of TUP, localities needed to be represented by street
centerlines. This way, dead spaces (non-populated areas) not identified by satellite images would
appear. However, in very small localities, a few TUP would cover the entire area, including the
dead spaces. Thus, it was also necessary to map street centerlines. A wide variety of land-base
maps were used.
Field Surveying
In addition to the maps, it was necessary to properly geo-reference TUP, regular schools and
health institutions. Computerized records existed with this basic information on TUP, and regular
school and health institution address databases were also available. Normally, this type of
information would reside in a computer mainframe database. However, complex logistics were
needed to manage the massive amount of changes. For instance, the number of TUP grew from
about 600,000 in 1998 to more than 1.5 million today. Most of this growth occurred in 2001.
TUP are not meant to generate profits. In some places, such as shopping centers, TUP are
heavily used and they are well protected from vandalism. In other areas, TUP are difficult to
maintain, but they play an important role in the region. TUP, in Brazil, use an inductive card
technology that provides proof of payment and are remotely supervised for accounting and
malfunctioning purposes. Several techniques can be used to survey TUP:
- Field surveying with GPS techniques
- Address interpolation using the TUP address
- Surveying the TUP and positioning it over a paper map for future digitalization
In the field, surveyors had to collect not only the coordinates of TUP placement but also
information on its type and features. The data collected included: address, infrastructure, 24-hour
availability, shelter type, access line identification, serial number, phone number, ability to
receive calls, accessibility, make/model, etc.
We estimate that the size of the survey crew exceeded 1,000. The TELCO managers and
subcontractors have a wealth of fascinating stories to tell about survey crew experiences. One
can imagine what could happen when surveying in slums, forests, during rainstorms, etc. At one
point, there was simply not enough precision GPS equipment available to equip all surveyors.
GIT
Each TELCO needed an AM/FM/GIS system to help manage this complex project. Spatial
databases were determined by region, but due to the administrative division of the concession
area, generally several of these regions had to be joined to allow for sophisticated queries and
algorithms making up the very large databases. However, in all cases, switching from one
database to another would have to be done within the same user session.
This is the basic mapping scenario:
- Division boundaries of the states inside the concession area
- Highways, roads, hydrography
- Division boundaries of the municipalities (counties) and their important features
- Division boundaries of the urban areas (localities) and their important features
- Aerial photos from diverse sources and different levels of precision
- Street centerlines and underline addressing databases
- Regular schools, health institutions
- TUP and TUP inductive card point of sales
- Other desirable objects input according to availability
As mentioned before, the TUP-MS required a high degree of availability, distribution over a
wide area network, and very important, usability. System management required use of Internet
access for ease of software deployment and maintainability. The use of modern GIT technology,
such as image compression, GPS, and graphic web access, were stretched to their limit.
As the maps were being readied, the already installed TUP had to be surveyed and properly
placed over them. This task required the use of several data conversion techniques. If a region
had good maps, with street centerlines, address ranges, and so on, an interpolation technique
could be used. This technique would get the TUP address from a conventional database and,
depending on the information available for the street centerlines, would interpolate the
geographic coordinates of the TUP placement inserting into the database the complete
information.
Imagine a database holding a complete map of the Sao Paulo city and surroundings (most of the
time, precise to the point of having each specific address geo-referenced) with its 10 million
inhabitants, including homes and office buildings. Now, imagine inserting by interpolation more
than 100,000 TUP. Then imagine having to do this more than once to accommodate updates and
adjustments. The algorithm would assume that there was a match for the TUP address in the
database. The TUP, normally, is placed in front of a regular address. If the maps were good
enough, this address would be already geo-referenced. If there was no match, the algorithm
would try to find the two addresses closest to the TUP address, on the same side of the street, to
get proper interpolation. This process would then try a series of situations until an interpolation
could be calculated even if there were no range addresses in the street segment. Of course, the
first attempt would verify if the street existed in the spatial database. In this case, and for the
other special cases, a detailed report was generated.
Another way to geo-reference the TUP was to get the TUP coordinates directly from GPS
devices. In this case, the information would be entered directly without the need for
interpolation. However, due to poor map precision, checks for closeness to street segments and
coordinates adjustments had to be performed. Still, another way to get the coordinates would be
to geo-reference the TUP position on top of a paper map and then obtain the coordinates with the
help of digitizing tablets.
After positioning the already installed TUP, and putting in place the update process for on-going
activities, the next step was to develop an optimizing algorithm to find the near optimal spots to
install the new TUP (or to move TUP positions) to comply with the regulation.
Let us assume that the optimal spots were identified for a given area. If the area had good maps,
plotting the proposed spots over the detailed map would be enough for the installation crew. The
installers would return the maps, and additional information on the TUP installed, to be
processed.
In fact, however, this process is quite complex, since it required integration with the legacy work
order systems already in use by the TELCOS. If the area had no street maps, the proposed spot
coordinates had to be transferred to GPS devices to help the installers. Several times, a proposed
spot would fall in a dead space (green areas, water reservoirs, etc) requiring special input and
new algorithm iterations to account for these situations. One of the TELCOS mastered this
technique to the extent that installers would find a solution, most of the time, in two iterations.
The algorithm for finding the near optimal spots for new TUP placements depends on the
method used to verify the goals. Remember the main goal was to have no user walk more than
300 meters to reach a TUP. A few techniques were used:
- From each TUP, paint the street segments paths up to 300 meters from the TUP. Then,
collect the street segments that were left partially or totally unpainted. This method required
the existence of very good street maps (and topological network) and was rarely used;
- From each TUP, paint a radius smaller than 300 meters (to account for walkers having to get
around the street blocks). Then, collect the areas not painted;
- Divide the area in sub-regions (squares, hexagons, etc.) of empirically determined sizes and
identify which sub-regions do not contain at least one TUP;
