The Conversion Model

As indicated earlier, one major goal of the GISOM project is the design of a cost-efficient conversion model for 7.5-minute quadrangles into DLG-3 files. The USGS started the conversion effort in 1978, and at the current conversion rate, it will take more than 300 years to complete the conversion for the whole country. In order to convert all of the 7.5' quadrangles in a reasonable number of years, a massive effort must be undertaken. Of course, conversion cost is a very important issue; high conversion process costs make it difficult to undertake a massive effort. Therefore, cost-efficiency is a major consideration in the design of the conversion model.

Great improvements in PC processing power and storage, together with the high degree of sophistication of PC-based CAD software, were very important factors in the development of the CFM conversion model. The current model uses an assortment of hardware and software with varying degrees of sophistication. Today, this model can be replicated using mostly PCs. The major components of the conversion model are described below.

The heart of the whole model is the scanning component. The high-resolution scanner and corresponding computer programs generate raster files. Any scanner with a resolution of at least 32 microns (800 dpi), which is the minimum resolution allowed by the USGS, can be used for the conversion of 7.5' quadrangles. The scanner is controlled by a computer and corresponding software. High-resolution scanners are driven today by PCs or workstations. Corresponding scanning software is available. The format of the output files depends on the scanner used. Today, file format is not a major problem, because there is enough information about most of these formats; therefore, it is relatively simple to convert files to the most appropriate format.

GISOM Conversion Sequence.

The scanning component is the most expensive part of the conversion model. Generally, a single high-resolution scanner can be used as part of several conversion efforts. If a scanner, similar to the one used in the GISOM Project, is used in two shifts (16 hours/day), it may be possible to scan eight quadrangles per day or 2080 quadrangles per year at 25 microns (1024 dpi) resolution (the resolution used in GISOM). Therefore, a single scanner driven by a PC could be used to support ten conversion projects similar to GISOM.

The other components of the conversion model are: warping, plotting, vectorizing and attributing, and QC/QA. Warping transforms scanned raster files into the map coordinate system. In general, scanning is performed in a local coordinate system similar to the local coordinate system of the hardcopy source document. In the GISOM project, the upper left corner is the local origin and the units are inches. Map features usually are given with respect to an absolute coordinate system. For the 7.5-minute quadrangles, the Universal Transverse Mercator map projection is used. In order to warp a raster image, two distinct operations need to be performed: a geometric transformation using common points and a resampling of the pixels to go from one space to the other.

Plotting generates hardcopies of raster files (after warping) and of vector files received from independent contractors. There are two components in plotting: a high-resolution plotter and the corresponding plotting software. Today, inexpensive ink-jet plotters driven by PCs perform this goal at a fraction of the cost of plotting with traditional electrostatic plotters.

Vectorizing and attributing are the components of the conversion model that generate graphic data in vector form with corresponding attributes. Three basic options for vectorizing exist: heads-up digitizing, interactive line following, and automated line following. Heads-up digitizing is the process of manually drawing vector cartographic features using the raster image as a backdrop. Interactive line following generates vector cartographic features by pointing to a line on the raster image and having a batch program following the line and generating its vector representation; automatic line following generates vector cartographic lines using a batch program that process all of them with a minimum user intervention. Attributing is the process of attaching, in an interactive or automatic fashion, a set of characteristics to the vector features, which complements their description and representation.

Figure 4. Conversion Model.

QC/QA is the component that controls the map conversion process to assure consistency and quality. Two major tasks of this component are conversion model evaluation and error detection. Conversion model evaluation studies the different components of the model and their interrelations and performance and makes decisions about their improvement. The error detection process finds and fixes positional, geometric, topological, and attribute problems in any layer. Figure 4 shows the main components of the conversion model.

The Conversion Model at The Ohio State University Center for Mapping

Scanning The Optronics 5040 is the scanner used in the GISOM Project. This is a raster scanner and film recorder device with an operating speed of 1 million pixels per second and the capability to scan color, as well as black and white data. It is a reflective drum scanner that accepts both reflective (opaque) and transparent media. Its maximum resolution is 2032 dpi (12.5 microns), and the maximum size of document to be scanned is 50’’ x 40’’. For the GISOM Project, each source document is scanned at 1016 dpi (25 microns) resolution to generate binary images. A set of predefined scanning parameters is used, and it takes about 20 minutes to scan each piece of source material (18’’ x 24’’).

Table 3. Scanning Parameters Used in GISOM. Scan units: English Media Classification: Mono Reflective Polarity: Negative Scan parameter file: Mono_rfl.scn Scan parameter units: Decimal Output format: Mono RL Media classification: Mono Reflective Media ID: Map... Scan resolution: 1016 DPI Media polarity: Negative Data format: Mono RLE Step mode: Continuous Scan region: Gain: 90 Min X: 2.00 Max X: 21.00 Threshold: 25/30 Min Y: 1.00 Max Y: 26.00 Aperture: Sharp 1016 DPI

The scanner is operated by SRIF (Intergraph's Scanner/Recorder Interface) software (UNIX-based). For the GISOM project, a set of standard parameters is used most of the time. Table 3 shows these values. Scanning is performed to each separate of a quadrangle (up to five separates per quadrangle), using the parameters above. The output format is a Run Length Encoded (.RLE) file (a binary scanning file). Size of files ranges from 1 or 2 megabytes to 30 or 40 megabytes. Raster files are plotted or displayed on the graphic screen to evaluate the quality of scanning. Three major factors are examined: resolution of the corner tic marks, location of the quadrilateral defined by the tic marks, with respect to the map features, and quality of line work. Tic marks must be about 7 pixels wide; map features must be completely inside the quadrilateral defined by the tic marks, and lines must be continuous. Also, contour lines should not touch each other. If problems are detected, the source material is scanned again, adjusting the parameters to try to overcome the problems. If better results cannot be obtained, new source material is requested from the USGS. The raster image is displayed using Intergraph I/RASB or BRAS.

Warping A set of programs developed in-house (CFM_ WARP) is used to warp the raster files. The process starts by generating a file that stores the coordinates of the corners of the quadrangle. These coordinates are expressed in Intergraph's unit of resolution (UOR). Then another program searches and automatically identifies the tic marks in the raster file (.RLE) generated by the scanner. Finally, another program uses an affine transformation to evaluate the accuracy of the tic control points. If they are satisfactory, an eight-parameter projective transformation, is used in combination with a backward resampling. The resulting raster file can be overlaid with an Intergraph design file containing the quadrangle corners. The warping software runs on the CFM mainframe (a Sun SPARCcenter 2000) for efficiency but could be moved to PCs.

Plotting The GISOM project currently uses an HP ink-jet plotter (DesignJet 650C). The plotting program used is CFM_PLOT, developed in-house, a PC-based program. Plotting the raster images is optional today. They are plotted only if requested by the independent contractors. Vector files are always plotted as part of the QC process.

Vectorizing and Attributing For vectorizing, only heads-up digitizing and automatic line following are operational in GISOM. Heads-up digitizing and attribute tagging involve manually selecting DLG-3 codes and tracing vector lines along a particular feature, using the raster image as a backdrop. Drawing is done on Intergraph’s design files (.DGN). PC-based software (RETSAM_PC_CFM) developed at the CFM is used for heads-up digitizing. RETSAM_PC_CFM has been developed, using Intergraph’s MicroStation as the drafting package, and it is based on the USGS workstation software RETSAM. Heads-up digitizing is used for all layers but hypsography. Automatic line following is used for the hypsographic layer.

After hypsographic files are warped, they are used as input to generate vector lines from a raster-to-vector program. A predefined set of rules for data classification, filter values to generate center lines, and tolerance values to close gaps and delete line spurs are used. The output of this process is a vector file representing the contours generated from the raster image in the coordinate space of the corresponding 7.5' quadrangle. The GISOM project has used the Intergraph I/VEC program for automatic line following. This is a UNIX-based program. An in-house contour vectorization program (described later) will replace the I/VEC program for Year 4 of the GISOM project.

The processes of contour correction and elevation and attribution tagging improve the vector lines generated by I/VEC and attach the corresponding DLG-3 codes (including the elevation) to each contour line. Attributes and elevations are stored in an ASCII file (.ATT). Two in-house programs, CONTOUR_EDITOR and TAGGER, are used for this operation. These programs run under MS-DOS on PCs.

QC/QA There are two major components for quality control: (1) geometric, topological, and attribute checking, and (2) visual checking. The PROSYS program developed by the USGS is the major tool used for geometric, topological, and attribute quality control. PROSYS runs every hour on the hour in a fully automated fashion on the CFM mainframe. This program requires the definition of a set of parameters. A file containing these parameters is available in the system. Some of these parameters include: map scale, quadrangle name, position of the southwest corner (latitude and longitude), and source date. The following PROSYS functions are used at the CFM: LNA (Line-Node Association) to create basic topology; CIP (Check Intersection) to check for valid line intersections (intersections only at end points) and for proximity of lines that do not actually intersect; LAA (Line-Area Association) to establish topological links between lines and areas; LHD (List Header Information) to list the DCF header; BOUPRO to check map elements and neat lines; STRCOP, DFN, DFA, and RUN to eliminate floating nodes, areas, and unnecessary nodes from the DCF files, and END to save all changes made to the DCF files.

The second major component of QC/QA is visual quality control. Raster images and hardcopies of quadrangles are used to compare the digital vector file and its attributes with the original map. Position errors, attribute errors, and errors of omission and commission are the main focus of this activity.

After quality control is performed in the files, error correction of line work and attributes (including elevation) is done in interactive fashion by the independent contractors using the RETSAM_ PC_CFM software. Figure 5 shows the production model implemented at the CFM.

Figure 5. GISOM Production Model.

Practical Implementation of the Conversion Model as Part of the GISOM Project

The following sequence of steps are followed in the GISOM project as part of the conversion of 7.5’ quadrangles to DLG-3 files. Steps 2-8 are part of the convertion model:

(1) Select the highest prioritized quadrangles. (2) Request and receive source material: evaluate material. (3) Scan source materials and evaluate raster file. (4) Warp raster files. (5) Plot raster files (optional). (6) Generate vector data and DLG-3 coding. (7) Plot vector files. (8) Quality Assurance (QA)/Quality Control (QC) project follow-up. If QA/QC is satisfactory, (9) Send DLG-3 data to USGS. If QA/QC is not satisfactory, (9’) Correct data and go back to (7) or (8). (10) QA/QC by USGS. If QA/QC is not satisfactory, go back to (9’).

The CFM, USGS, and independent contractors are responsible for different steps of the work described above. Figure 6 shows the distribution of the work.

Figure 6. GISOM Tasks Distribution.

The CFM’s tasks in the GISOM project are as follows: A senior project leader evaluates, scans, warps, plots, creates vector files, vectorizes, allocates work to the independent contractors, interacts with the state agencies and the USGS, and sets quadrangle priorities. This work is done with the cooperation of a student employee. Another senior project leader is responsible for testing new software tools, supervising and providing technical support (including software and hardware installation) to the independent contractors, and supervising the budget and accounting activities. Finally, another senior project leader is responsible for coordinating quality control activities, evaluating hypsographic files, and issuing payment orders for independent contractors. Hypsographic evaluation is done with the cooperation of a junior project leader. All of the above personnel are involved in quality control of nonhypsographic files, in cooperation with a highly skilled independent contractor.

Table 4. Tools Used by The Ohio State University Center for Mapping.

Project follow-up is done using ESRI’s ARC/ INFO as the basic software. The TRACKER application has been developed at the CFM to generate reports and graphics (including maps) that show the status of the project at any moment. Most of the GISOM staff is involved in updating the database and generating reports and maps from this application.

Table 4 shows in greater detail the tools used by The Ohio State University in performing the corresponding tasks.