Vol. VIII, No. 4

February, 1996

Drawing Profiles - Another Method

by H. Eiteljorg, II

In past issues of the CSA Newsletter I have discussed ways to speed the making of finished pottery profile drawings from pencil drawings made in the field (see "AutoCAD for Pottery Profiles," February, 1994, pp. 5-9, and "Pottery Profiles - Time and Costs," May, 1994, p. 5). (1) The first experiments were carried out at Carleton College in Minnesota in 1994, and the process was formalized at the CSA offices after my return.

Meanwhile, Professor Nancy Wilkie pursued another path, experimenting with a program named Draftsman®. Draftsman is a program designed to convert raster images to vector images automatically, and using it, Ms. Wilkie's students were able to automate the conversion of pencil drawings to computer profiles more fully. Professor Wilkie found that Draftsman could accurately trace the contour of a pot as drawn in pencil and scanned into a computer file, turning the raster image from the scanner into a vector image. The resulting drawing of the profile could then be used in any one of a number of programs.

In the fall of 1995, CSA obtained a copy of Draftsman Plus from the manufacturer, Arbor Image Corporation (5651 Plymouth Road, Ann Arbor, MI 48105 - 313-741-8700) to whom we express our thanks, for an expriment with Ms. Wilkie's system. Draftsman Plus operates under Microsoft Windows.

I began to use the program with Bryn Mawr undergraduate student, Alexis Menten, who was working on a project for Professor J. P. Dessel. A large number of sherd drawings was to be converted, and Ms. Menten and I experimented with the scanning process, the program for accepting and storing the scanned images (Adobe Photoshop[TM]), and the raster-to-vector conversion program, Draftsman.

One problem was encountered very quickly, because the original drawings had been done on graph paper. The graph lines needed to be removed, leaving only the profile to convert. It was not clear how the lines could be suppressed, and the seemingly obvious solutions did not work. Eventually it proved to be possible to remove the graph lines from the image, but the procedure for removing blue was different from that for removing orange (paper with lines of each color had been used in the field).

The original drawings were made with pencils and often had rather sketchy lines; so some clean-up was occasionally necessary to remove confusing lines. Good, clean drawings could be scanned and converted very easily and with almost no intervention, but originals with many erasures and accidental lines were more difficult and took longer to convert.

Conversion of the drawings with Draftsman was a straight-forward process, and it seemed promising as an alternative to tracing profiles. I decided to test this method, again using the pencil sketch from the 1994 AutoCAD experiment as the starting point. (Actually, the original sketch was lost in an office flood last summer; so I traced the drawing in the CSA Newsletter for February, 1994, in pencil and used the tracing as the starting point for this experiment.) The drawing was scanned into a Macintosh (the CSA scanner is connected to the MAC but could as easily be on a PC) using Adobe Photoshop. It was scanned as a gray-scale image. (Color would have been required, had the drawing been done on graph paper, since only the colors of the original lines on the graph paper make it possible to eliminate those lines efficiently.) The gray-scale image was converted to a black-and-white image; the Draftsman program requires black-and-white; no colors or grays are permitted. The resulting drawing was then saved to disk on the MAC and copied to the PC.

The drawing did not require much clean-up of extra lines or smudges; that process, had it been required, would have added to the time needed. So would the elimination of graph-paper lines, had the drawing been made on graph paper. (The added time would have been minimal for removing the lines.). The scanning required about five minutes. In a production setting, with many scans being made in succession, the time required would be about half that.

The scanned image was then loaded into Draftsman, and the program was instructed to trace the lines and to convert them from isolated dots into lines and arcs (vectors). Where necessary, the lines were adjusted. Stray lines were removed, lines were connected in places, and curves were modified. Since the new lines were shown directly on top of the original scanned image, they could be directly compared to the scanned version as they were adjusted. That was a simple and straight-forward process, and the program was even designed to permit the color of the scanned image to be adjusted so as to make the comparison easier.

The entire process to this point took about ten to twelve minutes, but making the separate lines into continuous lines for the pot and the handle (necessary to permit filling the outlined profile with black) added a good bit of time. The total was just over eighteen minutes. It then took another minute to add the black fill to the two shapes, and, since the program permits the use of drawing layers, the fill was placed on a separate layer. The time requirements stated were those of the fourth or fifth attempt with this drawing; by then I had had enough practice that I had achieved reasonable efficiency. The final requirement was to mirror the profile (without the fill) and draw the appropriate lines indicating exterior markings. The profile also was trimmed so that only the exterior remained on the mirror-image side. This added about ten minutes to the job. Printing took two minutes more.

After finishing the line drawing, the file was saved - and saved also as a CAD file (.dxf format for AutoCAD and also many other programs). Draftsman can save drawings in several other formats as well.

Printed versions of the drawing were compared to the original, and, as a result, the first drawing was modified to improve its appearance. (The time to do that was not included in the calculations, on the assumption that experience would make such modifications unnecessary.)

The results of this experiment were very interesting. First, the time required to get a complete drawing of the quality shown here in Fig. 1 (along with the AutoCAD drawing, the Illustrator drawing, and the artist's drawing from the original experiment) is on a par with the time required for other methods - thirty-one-and-one-half minutes for this method, twenty-nine minutes using AutoCAD, and thirty minutes for a hand drawing.

Profile created with Draftsman Profile created with AutoCAD and transferred to Illustrator Profile created with AutoCAD alone Profile created by artist

Second, continuous curves generated by AutoCAD from individual points along a profile (called splines) have been used in previous experiments with AutoCAD. Such machine-generated curves permit fewer data points to be used on long, continuous curves, and the result is a smoother curve than one obtains with many points along the same profile. To import a Draftsman-created profile into AutoCAD and add spline curves would require much more time and effort, but the AutoCAD drawing with those curves does seem to reflect the original better than the Draftsman version. Furthermore, the close inspection of the drawing involved in the experimental work showed that there were multiple lines in many places. Although they overlapped so as to be virtually indistinguishable, there were many unnecessary near-duplicate lines.

Third, using Draftsman allows the system operator to see clearly how well the original drawn shape has been copied; that is a significant advantage over the tracing method used in the experiment with AutoCAD. On the other hand, using AutoCAD directly makes it easier to rotate the shape around an axis or to determine volume and weight. (These things can be done by exporting the file to AutoCAD or another three-dimensional CAD program, but they can't be done within Draftsman.) Other differences in the way one works in the two programs may provide slight advantages in one system or the other, but they are not significant. Using AutoCAD probably requires more training.

The most important factor in the assessing these techniques has little or nothing to do with the drawing processes. That factor is how the results may be used by project personnel and by subsequent scholars. For project personnel, this means making it easy to produce plates for publication, whether for a paper or an electronic publication. For later scholars, the system used must make it easy to find, group, display, and print any number of profiles from a corpus. The system used should also make it easy to scale drawings accurately. One should be able to draw the profiles at any scale, with little or no difficulty, and to be certain that all profiles in a group have been drawn at the same scale.

There are two different approaches to storing and retrieving the drawings. One, based on the use of a CAD program, places all profiles in a single file, each on its own layer. (2) Using database-style layer names, it is possible to search for and select profiles according to any criteria designed into the layer-naming scheme. For instance, sherds from a particular site in a survey or of a particular period could be selected with a single command. The screen could then be divided so that many show at once, or the individual profiles could be moved to group them as the user wishes. There are several ways to accomplish this in a CAD program, but the important point is that the profiles from a project are all in a single file and accessible through a simple process of calling up layers. (Multiple CAD files may be viewed simultaneously; so profiles from different projects could be examined together.)

Producing scaled drawings, individually or in groups is a straight-forward process. Preparing them for publication may be somewhat more complex, but that process could be set up and repeated for different groups quite easily.

The other approach to storing and retrieving drawings is to place each profile in its own drawing file (using any of a number of programs, such as Draftsman itself, Illustrator[TM], CorelDraw[TM], etc.). Then one may combine many different profiles temporarily either in a single master drawing or in a desktop publishing package. Using one of these programs, for instance, one would simply open a file with a desired profile and then keep importing more files until all those profiles desired have been included. The same could be done with a desktop publishing program, and the user would have more flexibility in terms of the original file format, since desktop publishing packages usually accept many image formats. (3)

Of course, many profiles could be stored in a single drawing, but selecting the profiles required for a specific purpose would be nearly impossible.

Combining the profiles in a single drawing is time-consuming if not difficult. However, finding and selecting all those drawings that contain profiles meeting specific criteria is very difficult, especially if one is limited by the DOS eight-character limit for file names. Even if one could be certain that each profile had been found, perhaps with a separate database to provide a list of all selected examples, one would still need to bring each one into the master drawing individually. That would be very time-consuming.

Printing to scale from a graphics program can be somewhat more tricky than doing so with a CAD program, but, once mastered, the techniques are neither difficult nor problematic.

Familiarity with either process is likely to bias a user in favor of the familiar, but the advantages of using CAD seem to be significant to me. The operator has direct control over the selection of all points that define the shape of a sherd or vase. Long, continuous curves are better drawn with AutoCAD. Storing all the profiles from a single project in one AutoCAD model is more efficient. Making three-dimensional versions of vases, determining volume, and calculating weight require a CAD program or a modeling program. But AutoCAD is expensive ($4000 at list but about $1000 for academics), and a digitizer is also required (about $300).

Using Draftsman, on the other hand, permits the operator to compare directly the line of the profile to the original drawing during the work process. The program is also much less expensive than AutoCAD, but one also needs a scanner and a program like Photoshop to accept and modify the original scanned images. (There are three versions of Draftsman. The least expensive, at $339 will accept a 300 dots-per-inch scan from an image of a bit less than seven inches in width and any length; the mid-level version, at $995 will accept images up to a bit more than eleven inches in width and any length, the $1995 version will accept images of any size. There is a 50% academic discount. A scanner and image editing program will add about $1000 to $1500.)

Draftsman would be most useful on small sherds, particularly if they are rather irregular. For such items the use of long, continuous curves is minimal.

The major difference between these two approaches seems to be the system for storage and retrieval of the drawings. CAD systems with good layer-manipulation capabilities have a significant edge there. If a CAD system is to be used for storing and retrieving the drawings, then it seems more efficient to use the CAD system from the beginning.

For other Newsletter articles concerning pottery profiles and capacity calculations, consult the Subject index.

Next Article: Workshop for Surveying and AutoCAD

Table of Contents for the February, 1996 issue of the CSA Newsletter (Vol. 8, no. 4)

Table of Contents for all CSA Newsletter issues on the Web

Return to CSA Home Page

1. To convert a paper drawing of any kind into some form of computer drawing, one must either trace the sketch, as was done in the 1994 experiment, or scan the drawing electronically. The tracing creates a series of lines, arcs, and circles (collectively called vectors) in a scaled drawing. Such drawings are called vector images.

A scanned image, however, consists only of dots; the computer-jargon term for such images is raster images. Individual positions on the paper are read by the scanner, and each point has an assigned color value - or level of gray or simply black versus white value. Scanners can determine the color value of very small dots, but, no matter how small the dot, there are no true lines, only disconnected dots. In addition, lines on paper, even very fine ones, have width to a scanner so that the course of an individual line is somewhat imprecise in a scanned image. Its width will vary and will make the path somewhat ambiguous. Furthermore, there is no scale to such drawings, only a number of dots included in each direction, from which one may determine a scale in certain circumstances and with some difficulty. Therefore, many processes one might wish to apply - especially those involving measurements - are difficult with scanned images, and any conversion process from one file type to another carries with it the risk of losing references for scale. Return to text at note 1.

2. Layers, in CAD systems and other drawing programs, are simply individual groupings of items in the drawing. Items are simply drawn on different layers so that they can be suppressed in individual views or drawings. Return to text at note 2.

3. Draftsman and many of the drawing programs use layers; so one could place many profiles in a single drawing, each on its own layer. However, unlike some CAD programs, these programs do not include the layer-manipulation processes necessary to make the use of layers more than a drafting or printing aid. Return to text at note 3.