Raster to Vector Processing for Road Network Analysis

A vector database representing road networks can be used to derive information for a variety of purposes. In addition to its obvious utility in terms of describing the road network itself, this information can also be used as a means to characterize a range of variables. In many cases, the simplest means to obtain such a vector database is by extraction from an existing set of raster maps. Building the vector set from a raster source involves digitizing the road network, either by hand or with the assistance of an automated computer application. Once the vectors are stored in a convenient form, spatial operators are used to generate a parameter database describing the network. Those parameters can then be analyzed to yield information about population growth, land use conversion, change in impervious surface area and associated runoff, and several other indicators of development and change.

(top)

Return to Environmental Assessment Home

Because of recent decreases in the cost of disk space and processing power for computer

workstations, the collection and storage of large scale scanned raster maps is now practical for virtually all planning entities. When formatted in a common image file standard (TIFF, BMP, etc.), rasters are easy, cheap, and simple to store, access, view, transmit, and print. The information in these rasters, however, is frequently most efficiently utilized by a conversion to vector form whereby linear features can be stored as a digitized series of vector shapes. These vector files may be incorporated into a GIS where the points, lines, and areas are associated with a database of attributes, or they may be analyzed in vacuo to extract information intrinsic to the vectors themselves.

In order to extract linear elements from a raster map, some form of vector digitization must take place. Before the widespread availability of affordable scanning solutions, a human operator would use a paper map, a digitizing table, and a puck to place vector elements manually into a digital file or, in some instances, a GIS vector layer. Traditional hand digitizing, however, is being supplanted by more sophisticated techniques. Large format, high resolution scanners have obviated the necessity to deal with the paper map. "Heads up" digitizing capability in a variety of computer software applications has removed the digitizing table and puck from the workflow. Using the best of the current crop of dedicated raster to vector applications, linear map features can be extracted directly to vector form with only moderate intervention by human operators.

A vast array of tools is available for extracting vectors from a raster map. At the most basic end of the

spectrum are drawing tools such as Adobe Illustrator for PC or Aldus Freehand for Mac. The greatest advantage of these packages is their low cost. They have have vector tracing modules that are easy to use but are greatly limited by their "dumb" format. Vectors are stored without consideration of either geographic parameters (location of a vector on the surface of the earth or within the bounds of an area of interest) or for geodetic parameters (specific projections, ellipsoids, or datums). Additionally, the vectors produced by this sort of program has no provision for association with a database, rendering the output essentially unusable in a GIS.

A step above drawing tools are CAD programs. Applications such as Bentley MicroStation, Autodesk AutoCAD, and others have distinct advantages over drawing programs. Line creation tools specifically tailored for engineering-grade vector generation (arc segments, splines, automatic line extending and clipping, automated line cleaners and node weeders, etc.) greatly simplify the process of producing accurate vectors. In addition to this, vectors produced by CAD packages have built-in provision ("hooks") for association with databases. Often countering these advantages, though, are large size and daunting complexity. CAD programs are, at their core, multifaceted design tools and may contain utilities for electrical engineering, civil engineering, architecture, facility design, modeling, and other various disciplines in addition to whatever geographical and mapping functionality they provide. Because of this complexity, considerable training may be necessary for an operator to become productive and efficient with the program.

The most sophisticated tools for the job are purpose-built raster to vector applications. The most capable such as Z/I Imaging's Geovec Office and ESRI's Feature Analyst extension for ArcView and ArcGIS are highly refined suites of feature extraction tools with the capability to feed their vector output directly into existing GIS databases or export the data to industry standard file formats for further analysis or ingestion into other software packages. Automated and interactive utilities cleanly and efficiently produce intelligent vectors in geographically-aware workspace. The processes described below assume the use of a product at this level.

(top) Return to Environmental Assessment Home

In preparation for processing, paper maps should be scanned utilizing a large format, high resolution

scanner. A variety of suitable scanners are available from Contex, Vidar, Action Imaging, and others. When selecting a scanner for creating a raster data set, the primary considerations are the capability to handle the paper maps in a single scan, minimal positional distortion,and sufficient scan resolution. Resolution requirements vary by task. For simple archival scans, 200dpi is generally considered adequate. Vector output to GIS demands considerably more resolution, something on the order of 400dpi to 600dpi. For general raster to vector operations, 300dpi to 400dpi is normally adequate. Depending on the software used, there may or may not be an advantage in saving the raster in a color format, a grayscale format, or simple black and white. Refer to the manufacturer's documentation when selecting a storage format.

A typical extraction workflow begins by initializing the raster to vector software and opening the scanned map as a backdrop. In order for the line following algorithm to discern what regions of the raster constitute pixels representing roads, several parameters must be set within the configuration section of the software. The parameters serve to characterize the appearance and spatial

characteristics of the road network, including color, minimum and maximum road width, intersection angles, and several other descriptive aspects. In addition to the characterizing parameters, several decision switches and limits must be set. These decisions address situations in which, for instance, the algorithm encounters a fork in the road or a gap in an otherwise continuous line, as well as determining the direction in which the algorithm progresses along the raster path, the length of individual vector segments in curves, and others aspects of vector creation.

Once the parameters are defined, the algorithm is given a starting point using the cursor to select a point on a roadway. From that point, vectors are created along the raster path in a continuous chain and displayed on the computer monitor, pausing only for operator input where the preset parameters cannot adequately guide a decision, where paths are uncertain at intersections, or where a feature terminates. At a line terminus, the algorithm pauses for operator input which consists of selecting another start point for a vector chain or, if collection is complete for a particular section of the network, ending the creation process completely. When all segments of all aspects of the network have been collected, the vectors are typically inspected, cleaned, clipped, and/or extended where necessary, and stored to hard disk. The exact process of this editing is dependent upon the software being used and will normally be described in the software documentation.

(top) Return to Environmental Assessment Home

Introduction
Background
Vector Extraction
	Return to Environmental Assessment Home