Keywords: Neural Networks, Remote Sensing, GIS, Urban Transportation
Abstract It is described in this paper a formulation of a travel forecast model that
incorporates the geographical– spatial reality of urban areas. Throughout such incorporation, the
model aims to characterize the interaction between land use – transportation system, which is
the basis to represent the displacements (trips) needs of human activities within the urban area.
Geographical Information System, Remote Sensing are explored in the obtainment and
processing of geographical– spatial data, mainly originated from satellite and aerial photograph
images. The comprehension of urban dynamics that affect travel process is established through
an “artificial intelligent” approach using Neural Networks. A case study is conducted in order to
evaluate and discuss NN modelling for travel forecast.
1. Introduction
In order to analyze urban area problems, Remote Sensing (RS) and Geographical Information
Systems (GIS) integration has been successfully applied. Morain and Baros (1996) verified that
RS-GIS integration data has lead to time reduction and cost-saving benefits against field surveys
that are hardly updated by using traditional methods. However, Foresman and Millete (1997)
emphasize that GIS capabilities have been concentrated in boolean operations and conventional
spatial statistics when processing multilayered map-based information that has lead to a basic
level of “display analysis”.
Recently, many efforts are observed intending to achieve more powerful spatial analytic and
spatial process modelling tools by exploring Neural Networks (NN). According to Fischer
(1999), computational geographers have faced NN as an instrument to overcome limitations of
conventional tools to explore patterns and relationships in GIS and RS. Openshaw (1991) and
Openshaw et al (1995) reported successful applications and NN efficiency, showing
perspectives to elaborate a new modelling style. In this sense, it is essential to create models
concerning this potential when integrated to GIS and RS.
Such potential is decisive to provide an efficient representation of urban dynamic that directly
affects problems and solutions into the city. Specifically in transportation studies, travel demand
is a consequence of how urban activities are processed and organized. This process is
represented by urban dynamic along the years. Traditional forecast models quantify how urban
activities are converted into travels considering socioeconomic data (population, income, etc.).
Differing from previous formulations, this research proposes that travel forecast can be achieved
analyzing land use and transportation system interactions in a spatial-temporal evaluation. These
interactions could not be an easy task using traditional mathematical and statistical tools, but the
integration of RS, GIS and NN concentrates together elements that are capable to represent the
spatial characteristics of this problem. RS and GIS contribute to provide land use and
transportation system data/information, which are mainly collected from satellite images and
aerial photographs. On the other hand, NN gives support to the complex analysis of the
interactions along the time and space.
This paper emphasizes this integration and presents a preliminary formulation of travel forecast
model considering only the spatial dimension. Mainly, it is aimed to define basic characteristics
and NN structures that after will be part of the spatial-temporal model. These definitions play a
decisive role, since model’s essence will be established from these activities. Tests for Boston
Metropolitan Area were conducted in order to verify the efficiency using this type of networks.
2. RS and GIS and NN Integration Approach
This integration approach allows incorporating to travel forecast model spatial-temporal
characteristics related to urban area activities. It is necessary to develop a framework that
combines RS and GIS and NN potential, involving steps from the data collection until the
visualization and result analysis. In the development process, it is important to be concerned
about the capabilities of each integration element. In this section, the proposed integration was
created considering the whole process and specially NN definition for modelling in spatial
dimension. Then, it is focused on the process description emphasizing on the framework
formation in order to evaluate future development needs. Specific issues related to format,
resolution and sources integration are not treated here.
Five modules compose the integration, that are: RS data; Trips data; Maps data; GIS; and NN.
In figure 1, it can be verified that the process starts with data obtainment that is associated to the
correspondent module. Thus, satellite images and aerial photographs are stored in RS data
Module (1). The origin / destination tables are organized in Trips data Module (2) and finally
maps containing transportation system information (road, subway, train, bus, etc) are transferred
to Maps data Module (3). The data from Module 1 suffers a preliminary treatment inside GIS
Module (4), in order to process multispectral analysis and aerial photograph interpretation. From
this treatment, land use patterns are reached and gather together with data from Module 2 and 3
inside GIS database. Spatial-temporal queries are processed using this database and the results
are conducted to NN Module (5). In this module, query results are pre-processed and then
divided in training and test data sets. NN software performs the training process defining a
function that is tested and revised until reach the expected level of forecast. Finally, Module 5
returns to Module 4 (GIS), where travel forecast function is applied to a future scenario and then
visualized and analyzed.
