Home | Business News | Browse by Publication | C | Cartography and Geographic Information Science

A snake-based approach for TIGER road data conflation.

Article Excerpt
Introduction

The U.S. Census Bureau's TIGER database is primarily used to support the mapping geographic analysis and other GIS activities that serve the various censuses and surveys conducted by the Bureau. The initial sources used to create the database were the U.S. Geological Survey (USGS) l:100,000-scale Digital Line Graph (DLG), the U.S. Census Bureau's 1980 geographic base files (GBF/DIME-Files), and a variety of miscellaneous maps. In order to maintain a current geographic database, TIGER has been continuously updated using a wide variety of source materials and techniques. Some updates have come from map annotations made by enumerators as they attempted to locate living quarters by traversing every street feature in their assignment area. The Census Bureau digitized these updates directly into the TIGER database without geodetic controls or the use of aerial photography to confirm features' positional accuracies. Other corrections and updates were supplied by local participants in various Census Bureau programs. Maps were sent to participants for use in various programs, and some were returned with updated annotations and corrections. The Census Bureau generally added the updates to the TIGER database without extensive checks. Changes made by local officials did not have geodetic control (Grady and Godwin 2000).

Whereas relational accuracy once was adequate for census activities, changing goals and technologies now require that the Census Bureau improve the positional accuracy of TIGER. One of the Bureau's goals for the next decade is to capture the latitude and longitude coordinates for living quarters and to provide each field interviewer with portable computers equipped with Global Positioning System (GPS) technology to accurately locate living quarters requiring a visit. However, to integrate the more accurate coordinates that GPS can provide for living quarters (and the streets they are along) with the existing MAF (master address file) / TIGER System, current TIGER features must have an equivalent level of positional accuracy. Additionally, the Census Bureau has found that many local GIS files have a greater positional accuracy; hence, the current positional accuracy of TIGER is a limiting factor. It precludes more effective address lists and geographic information partnerships with those state, local, and tribal governments that have high-quality address, street, boundary, and related geographic information. The successful partnerships with state, local, and tribal agencies rely upon improving the positional accuracy of TIGER (Marx 2000).

The Census Bureau launched the MAF/TIGER enhancement program in 2000. The first strategic objective was to improve address/street location accuracy and implement automated change detection. The Census Bureau sought to achieve a high level of map coordinate accuracy in TIGER by acquiring and using, as a first priority among data sources, digital files prepared and provided by state, local, and tribal governments (Broome and Godwin 2003).

Conflation approaches are often used to improve TIGER road data. However, this traditional approach requires two vector datasets. It does not make use of the high resolution imagery that is often readily available and more current. In addition, to deal with non one-to-one feature matching, the splitting of road line segments or attributes is often performed manually by a human operator, which is labor intensive and time consuming.

In this paper we propose a new hybrid conflation approach that combines traditional conflation with active contour models (snakes). Snakes are deformable contours that can move iteratively under the influence of internal forces and external image forces. The basic processes are feature matching, map alignment, and position correction by snakes. The next section presents an overview of traditional conflation methodology. This is followed by our snake-based conflation approach. Experimental results are then presented, followed by discussions and conclusions.

Traditional Conflation Methodology

Although local GIS data may have better positional accuracy than TIGER data, most do not have richer attributes. To get the best spatial and attribute information from both sources, conflation technology was developed. Conflation is the process that combines two spatial datasets of the same region to produce a superior dataset that is better than either source in spatial and attribute aspects. Through the conflation process, individual strengths of the source datasets can be combined. A dataset with excellent spatial accuracy but little attribute information can be merged with one with rich attribute information but poor spatial accuracy to produce a new dataset that is both spatially accurate and attribute rich.

The history of conflation goes back to the early 1980s, when the first development and application of an automated conflation process occurred during a joint U. S. Geological Survey--U.S. Census Bureau project designed to integrate the agencies' respective digital map files of U.S. metropolitan areas (Saalfeld 1988). The implementation of a computerized system for this task provided an essential foundation for much of the theory...

View this article FREE - Now for a Limited Time, try Goliath Business News
Free for 3 Days!